CN115438443A - Bearing residual life prediction method, device and system and readable storage medium - Google Patents
Bearing residual life prediction method, device and system and readable storage medium Download PDFInfo
- Publication number
- CN115438443A CN115438443A CN202211098106.0A CN202211098106A CN115438443A CN 115438443 A CN115438443 A CN 115438443A CN 202211098106 A CN202211098106 A CN 202211098106A CN 115438443 A CN115438443 A CN 115438443A
- Authority
- CN
- China
- Prior art keywords
- bearing
- time domain
- residual
- detected
- life
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/17—Mechanical parametric or variational design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Computer Hardware Design (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Computational Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The application provides a method, a device and a system for predicting the residual life of a bearing and a readable storage medium. The method comprises the steps of obtaining time sequence data to be measured of a bearing to be predicted; extracting the time domain characteristics to be detected of the time sequence data to be detected; obtaining an estimated value associated with the time domain feature to be detected by utilizing a time sequence processing model according to the time domain feature to be detected, wherein the time sequence processing model is obtained by training full-period data of a bearing sample, the full-period data of the bearing sample comprises health data of the bearing sample and fault data of the bearing sample, and the estimated value is obtained by utilizing the time sequence processing model according to the time domain feature to be detected and the health data of the bearing sample; determining a residual value according to the estimated value and the time domain feature to be detected; and determining the bearing life corresponding to the residual value according to the one-to-one corresponding relation between the residual value and the bearing life, and taking the bearing life corresponding to the residual value as the residual life of the bearing to be predicted. In this way, the predicted bearing life is more accurate.
Description
Technical Field
The present application relates to the field of bearing detection technologies, and in particular, to a method, an apparatus, a system, and a readable storage medium for predicting a remaining life of a bearing.
Background
The bearing is an important transmission component of mechanical equipment, and the working condition of the bearing can have great influence on the equipment. The failure of the bearing often reduces the reliability and precision of the equipment, which not only affects production and reduces the service life of the equipment, but also causes accidents. Therefore, the prediction research on the residual service life of the bearing is significant.
In the prior art, RNN is used for processing a time sequence signal of a bearing and predicting the residual life of the bearing. Due to the short cycle of RNN learning data, such as one month of data, only one month of data can be predicted, and for longer time of data, gradient disappearance and gradient explosion occur, resulting in inaccurate predicted bearing life.
Disclosure of Invention
The application provides a method, a device and a system for predicting the residual life of a bearing and a readable storage medium, and the predicted bearing life of the method is more accurate.
The application provides a method for predicting the residual life of a bearing, which comprises the following steps:
acquiring time sequence data to be measured of a bearing to be predicted;
extracting the time domain characteristics to be detected of the time sequence data to be detected;
obtaining an estimated value associated with the time domain feature to be detected by using a time sequence processing model according to the time domain feature to be detected, wherein the time sequence processing model is obtained by training full-period data of a bearing sample, the full-period data of the bearing sample comprises health data of the bearing sample and fault data of the bearing sample, and the estimated value is obtained by using the time sequence processing model according to the time domain feature to be detected and the health data of the bearing sample;
determining a residual value according to the estimated value and the time domain feature to be detected;
and determining the bearing life corresponding to the residual value according to the one-to-one corresponding relation between the residual value and the bearing life, and taking the bearing life corresponding to the residual value as the residual life of the bearing to be predicted.
Further, the method for predicting the residual life of the bearing further comprises the following steps:
acquiring full-period data of a bearing sample;
extracting time domain characteristics of the full-period data;
processing the time domain features to obtain the processed time domain features, wherein the processed time domain features comprise time domain features corresponding to the health data and time domain features corresponding to the fault data;
training the time sequence processing model by using the processed time domain characteristics;
inputting the time domain characteristics corresponding to the health data into the time sequence processing model to obtain a predicted value associated with the time domain characteristics corresponding to the health data;
determining a residual curve between the predicted value and the time domain characteristic corresponding to the fault data, wherein the residual curve comprises a residual value of the fault data; the fault data comprises the fault occurrence starting time and the fault occurrence time period of the fault data;
and determining a one-to-one corresponding relation between the residual value and the service life of the bearing according to the residual curve.
Further, the processing the time domain feature to obtain the processed time domain feature includes:
normalizing the time domain characteristics to obtain the normalized time domain characteristics;
the training of the time sequence processing model by using the processed time domain features and the obtaining of the predicted value associated with the time domain features to be tested comprise:
and training the time sequence processing model by using the normalized time domain characteristics.
Further, the method further comprises:
acquiring a one-to-one corresponding relation between the residual value and the service life of the bearing by adopting the following method:
constructing a residual curve according to the residual data of the bearing sample and the bearing life of the bearing sample;
and fitting the residual error curve to obtain a fitted curve, wherein the residual error curve comprises a one-to-many corresponding relation between the residual error value and the service life of the bearing, and the fitted curve comprises a one-to-one corresponding relation between the residual error value and the service life of the bearing.
Further, the fitting the residual error curve to obtain a fitted curve includes:
and fitting a residual error curve corresponding to the fault occurrence time period by using an exponential function to obtain a fitted curve, wherein the residual error curve comprises a one-to-many corresponding relation between residual error values and bearing service lives, and the fitted curve comprises a one-to-one corresponding relation between the residual error values and the bearing service lives.
Further, the time sequence processing model comprises a Bayes-LSTM model or a Bayes-GRU model.
Further, obtaining an estimated value associated with the time domain feature to be detected by using a time sequence processing model according to the time domain feature to be detected includes:
processing the time domain feature to be detected to obtain the processed time domain feature to be detected;
and inputting the processed time domain feature to be detected into the time sequence processing model to output the estimated value.
Further, the processing the time domain feature to be detected to obtain the processed time domain feature to be detected includes: selecting any time domain feature to be detected representing the degradation trend of the bearing from the time domain features to be detected of the time sequence data to be detected; normalizing any selected time domain feature to be detected to obtain the normalized time domain feature to be detected; the inputting the processed time domain feature to be detected into the time sequence processing model to output the estimated value includes: inputting the normalized time domain feature to be detected into the time sequence processing model to output the estimated value; determining a residual value according to the estimated value and the time domain feature to be detected, wherein the determining comprises the following steps: determining a residual error between the estimated value and any selected time domain feature to be detected as the residual error value;
and/or the presence of a gas in the gas,
the processing the time domain feature to be detected to obtain the normalized time domain feature to be detected includes: selecting a plurality of time domain characteristics to be tested representing the degradation trend of the bearing from the time domain characteristics to be tested of the time sequence data to be tested; normalizing the selected multiple time domain features to be detected to obtain normalized time domain features to be detected; the inputting the processed time domain feature to be detected into the time sequence processing model to output the estimated value includes: inputting the time domain feature to be detected after normalization processing into the time sequence processing model to output the estimated value; determining a residual value according to the estimated value and the time domain feature to be detected comprises: determining a plurality of residual errors between the estimated value and the selected plurality of time domain features to be detected; determining an average value of the plurality of residuals as the residual value.
The application provides a bearing residual life prediction device, includes:
the acquiring module is used for acquiring to-be-predicted time sequence data of the bearing to be predicted;
the extraction module is used for extracting the time domain characteristics to be detected of the time sequence data to be detected;
the time domain characteristic measuring device comprises a first processing module, a time sequence processing module and a second processing module, wherein the first processing module is used for obtaining an estimated value related to the time domain characteristic to be measured by utilizing the time sequence processing model according to the time domain characteristic to be measured, the time sequence processing model is obtained by training full-period data of a bearing sample, the full-period data of the bearing sample comprises health data of the bearing sample and fault data of the bearing sample, and the estimated value is obtained by utilizing the time sequence processing model according to the time domain characteristic to be measured and the health data of the bearing sample;
the second processing module is used for determining a residual value according to the estimated value and the time domain feature to be detected;
and the third processing module is used for determining the bearing life corresponding to the residual value according to the one-to-one corresponding relation between the residual value and the bearing life, and the bearing life is used as the residual life of the bearing to be predicted.
The application provides a bearing remaining life prediction system, which comprises one or more processors and is used for realizing the method in any one of the above.
The present application provides a computer readable storage medium having stored thereon a program which, when executed by a processor, implements a method as described in any one of the above.
In some embodiments, the method for predicting the residual life of the bearing uses a time sequence processing model trained by the full life cycle data of the bearing sample according to the time domain feature to be detected to output an estimated value associated with the time domain feature to be detected; determining a residual value according to the estimated value and the time domain characteristic to be detected; and determining the bearing life corresponding to the residual value as the residual life of the bearing to be predicted according to a one-to-one corresponding relation between the residual value and the bearing life, wherein the time sequence processing model is obtained by training full-life-cycle data of a bearing sample, the full-life-cycle data of the bearing sample comprises health data of the bearing in a healthy state and fault data of the bearing in a fault state, and the estimated value is obtained by using the time sequence processing model according to the time domain characteristics to be measured and the health data of the bearing sample. Therefore, according to the time domain characteristics to be predicted of the bearing, the time sequence processing model trained by the full life cycle data of the bearing sample is used for outputting the estimated value associated with the time domain characteristics to be predicted, and finally the bearing life corresponding to the determined residual value is obtained and used as the residual life of the bearing to be predicted. Thus, the time sequence processing model can process time sequence data for a longer time, so that the predicted residual service life of the bearing to be predicted is more accurate.
Drawings
Fig. 1 is a schematic structural diagram of a wind turbine generator according to an embodiment of the present application;
fig. 2 is a schematic flowchart illustrating a method for predicting the remaining life of a bearing according to an embodiment of the present disclosure;
FIG. 3 is a schematic diagram illustrating time-domain feature RMS data taken as an example of a full life cycle of a bearing according to an embodiment of the present application;
FIG. 4 is a graph illustrating the prediction of bearing RMS data in the method of predicting the remaining life of a bearing shown in FIG. 2;
FIG. 5 is a graph showing a residual error curve in the method for predicting the residual life of the bearing shown in FIG. 2;
FIG. 6 is a schematic flow chart illustrating the steps 130 and 140 of the method for predicting the remaining life of the bearing shown in FIG. 2;
FIG. 7 is a schematic diagram showing another detailed flowchart of steps 130 and 140 in the method for predicting the remaining life of a bearing shown in FIG. 2;
FIG. 8 is a diagram illustrating an example of applying the peak-to-valley values in the time domain characteristic of the method for predicting the remaining life of the bearing shown in FIG. 2;
FIG. 9 is a schematic diagram illustrating an example of the RMS application of the time domain characteristics of the method of predicting the remaining life of a bearing shown in FIG. 2;
FIG. 10 is a diagram illustrating an example of an application of the pulse factor in the time domain characteristic of the method for predicting the residual life of the bearing shown in FIG. 2;
FIG. 11 is a diagram illustrating an example of an application of kurtosis in a temporal signature of the bearing remaining life prediction method shown in FIG. 2;
FIG. 12 is a schematic diagram illustrating an example of the residual error sum curve and the fitting curve thereof in the method for predicting the residual life of the bearing shown in FIG. 2;
FIG. 13 is a schematic view showing an example of application of the residual life of the bearing in the method for predicting the residual life of the bearing shown in FIG. 2;
FIG. 14 is a block diagram illustrating a residual life predicting apparatus for a bearing according to an embodiment of the present invention;
fig. 15 is a block diagram illustrating a structure of a system for predicting remaining life of a bearing according to an embodiment of the present application.
Detailed Description
Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the exemplary embodiments below do not represent all embodiments consistent with one or more embodiments of the specification. Rather, they are merely examples of apparatus and methods consistent with certain aspects of one or more embodiments of the specification, as detailed in the claims which follow.
It should be noted that: in other embodiments, the steps of the corresponding methods are not necessarily performed in the order shown and described herein. In some other embodiments, the method may include more or fewer steps than those described herein. Moreover, a single step described in this specification may be broken down into multiple steps in other embodiments; multiple steps described in this specification may be combined into a single step in other embodiments.
Fig. 1 is a schematic structural diagram of a wind turbine generator 10 according to an embodiment of the present application. As shown in fig. 1, the wind turbine 10 includes a tower 12 extending from a support surface 11, a nacelle 13 mounted on the tower 12, and a rotor 14 assembled to the nacelle 13. Wind rotor 14 includes a rotatable hub 15 and at least one blade 16, blade 16 being connected to hub 15 and extending outwardly from hub 15. In the embodiment shown in fig. 1, the wind rotor 14 includes three blades 16. In some other embodiments, the wind rotor 14 may include more or fewer blades 16. A plurality of blades 16 may be spaced about hub 15 to facilitate rotating wind rotor 14 to enable wind energy to be converted into usable mechanical energy, and subsequently, electrical energy.
The wind turbine 10 further comprises bearings (not shown). The bearings (not shown) include a yaw bearing (not shown) and a pitch bearing (not shown). A yaw bearing is mounted at the junction of the tower 12 and the nacelle 13, the yaw bearing acting as a sliding bearing between the nacelle and the tower for transferring forces from the wind turbine 10 to the tower 12. A pitch bearing (not shown) is mounted at the junction of the root of each blade 16 and the hub 15. The pitch bearing is used for changing the attack angle of airflow to the blade 16 by adjusting the pitch of the blade 16 when the wind speed is too high or too low, so that the aerodynamic torque obtained by the wind turbine generator 10 is changed, and the power output is kept stable. Wherein the pitch bearing (not shown) and yaw bearing (not shown) for the pitch and yaw system are oversized slewing bearings.
Of course, the bearings (not shown) may also be installed on the rotor main shaft (not shown), the gearbox such as a speed increaser (not shown), the generator (not shown), the yaw gearbox such as a speed reducer (not shown), the yaw rotating base (not shown), the pitch rotating base (not shown) of the blades 16, the hydraulic pump (not shown), and the like of the wind turbine generator 10. This is not exemplified.
In order to solve the technical problem that the predicted bearing life is inaccurate, the embodiment of the application provides a method for predicting the residual life of a bearing, wherein a time sequence processing model trained by full life cycle data of a bearing sample is used according to the time domain feature to be predicted of the bearing to output an estimated value associated with the time domain feature to be predicted; determining a residual value according to the estimated value and the time domain characteristic to be detected; and determining the bearing life corresponding to the residual value as the residual life of the bearing to be predicted according to a one-to-one corresponding relation between the residual value and the bearing life, wherein the time sequence processing model is obtained by training full-life-cycle data of the bearing sample, the full-life-cycle data of the bearing sample comprises health data of the bearing sample in a healthy state and fault data of the bearing sample in a fault state, and the estimated value is obtained by using the time sequence processing model according to the time domain characteristics to be predicted and the health data of the bearing sample. Therefore, according to the time domain characteristics to be predicted of the bearing, the time sequence processing model trained by the full life cycle data of the bearing sample is used to output the estimated value associated with the time domain characteristics to be predicted, and finally the bearing life corresponding to the determined residual value is obtained and used as the residual life of the bearing to be predicted. In this way, the time sequence processing model can process time sequence data for a longer time, so that the predicted residual service life of the bearing to be predicted is more accurate.
Fig. 2 is a schematic flow chart of a method for predicting the residual life of a bearing according to an embodiment of the present application.
As shown in fig. 2, a method for predicting the remaining life of a bearing provided by an embodiment of the present application includes the following steps 110 to 150.
And step 110, acquiring time sequence data to be measured of the bearing to be predicted.
The bearing to be predicted is used for reflecting the information required to be detected by the bearing. And the time sequence data to be measured is used for reflecting the time sequence data in the running process of the bearing to be predicted. Here, the bearing to be predicted and the bearing sample mentioned below are used to distinguish between the two bearings, and different names are used. The bearings may include, but are not limited to, sliding bearings and/or rolling bearings. The rolling bearing is an important rotating part in mechanical equipment and is one of important failure sources of the mechanical equipment. Compared with other mechanical parts, the bearing has the great characteristic that the service life of the bearing is very discrete. Some bearings have been greatly beyond the design life yet still perform well, while some have failed well short of the design life. Therefore, the service life of the bearing can be accurately predicted, sudden shutdown and accidents caused by bearing faults can be avoided, and the service time of the bearing can be maximized.
And step 120, extracting the time domain characteristics to be detected of the time sequence data to be detected.
When the bearing to be predicted breaks down, the characteristic values of the time domain waveform of the bearing to be predicted change, and some characteristic values can represent the degradation trend of the bearing and can be used for predicting the residual life of the bearing to be predicted. The time domain characteristics to be measured are used for reflecting the data of the time domain characteristics of the time sequence data to be measured of the bearing to be predicted. These temporal features may include, but are not limited to, one or more of a peak-to-valley value, a Root-Mean-Square (RMS) value, a pulse factor, and a kurtosis.
Wherein the peak-to-valley value is used to reflect the instantaneous intensity of the reaction force signal impact. The peak-to-valley value represents the change of signal intensity, and is commonly used for judging impact faults. The root mean square value (effective value) RMS is used to reflect the magnitude of the signal's vibrational energy. The pulse factor is used to detect an indication of whether there is an impact in the signal. Kurtosis is used to indicate how flat a waveform is, and is used to describe the distribution of variables. The kurtosis of a normal distribution is equal to 3, the curve of the distribution is "flat" for kurtosis less than 3 and "steep" for a distribution greater than 3. Response to vibration signal shock characteristics. The kurtosis is the most common characteristic index of pitting damage faults, and is particularly suitable for monitoring early faults of rolling bearings. Kurtosis is insensitive to wear-type failures. In other embodiments of the present application, the time series data signal of the bearing is not limited to the stress signal, and may be a displacement signal, an acceleration signal, a strain signal, or the like as long as the operating state of the bearing can be reflected.
Time series data x = { x 1 ,x 2 ,…,x n ,…,x N }
And step 130, obtaining an estimated value associated with the time domain feature to be detected by using a time sequence processing model according to the time domain feature to be detected, wherein the time sequence processing model is obtained by training full-life-cycle data of a bearing sample, the full-life-cycle data of the bearing sample comprises health data of the bearing sample and fault data of the bearing sample, and the estimated value is obtained by using the time sequence processing model according to the time domain feature to be detected and the health data of the bearing sample. Wherein, the health data of the bearing sample is the data in a healthy state. The fault data of the bearing sample is the data when the bearing sample is in a fault state. Fig. 3 shows the full life cycle data of the bearing sample, and fig. 3 is a schematic diagram of the full life cycle data of the bearing sample according to the embodiment of the present application, which is exemplified by time-domain feature RMS data.
The time sequence processing model comprises a Bayes-LSTM (Long Short-Term Memory) model or a Bayes-GRU (Gated Recurrent Unit) model. The Bayes-LSTM model or Bayes-GRU model can be trained and predicted for multiple times on the same data respectively, and the training results of multiple times are averaged to output the estimation value corresponding to the processed time domain feature. Therefore, bayes is used for carrying out uncertainty expression, and multiple training results are averaged, so that deviation data is removed, errors are blurred, and the accuracy of output values is improved.
The step 130 may further include two steps, a first step of processing the time domain feature to be detected to obtain a processed time domain feature to be detected; and a second step of inputting the processed time domain feature to be detected into the time sequence processing model to output an estimated value. The processed time sequence feature to be detected refers to the real value of the processed time sequence feature to be detected. Therefore, the time domain characteristics to be detected are processed, the complexity of input data is reduced, and the data processing efficiency is improved. The processing includes selection and/or normalization. The details are as follows.
Wherein, the time sequence processing model is obtained by training the full life cycle data of the bearing sample, and the following 7 steps are further explained: the method for predicting the residual life of the bearing further comprises the following 4 steps: step 1, acquiring full life cycle data of a bearing sample. And 2, extracting time domain characteristics of the full life cycle data. And 3, processing the time domain characteristics to obtain processed time domain characteristics, wherein the processed time domain characteristics comprise time domain characteristics corresponding to the health data and time domain characteristics corresponding to the fault data. And 4, training the time sequence processing model by using the processed time domain characteristics. And 5, inputting the time domain characteristics corresponding to the health data into the time sequence processing model to obtain a predicted value associated with the time domain characteristics corresponding to the health data. The term is used for distinguishing, the term is output when the time sequence processing model is based on the full-period data training of the bearing sample and is called a predicted value, and the term is output when the time sequence processing model is used based on the time sequence data to be predicted of the bearing to be predicted and is called an estimated value. As shown in fig. 4, fig. 4 is a diagram illustrating an example of the prediction of the residual life of the bearing shown in fig. 2, which is based on the RMS data of the bearing. Step 6, determining a residual error curve between the predicted value and the time domain characteristic corresponding to the fault data, wherein the residual error curve comprises a residual error value of the fault data; the failure data includes a failure occurrence start time and a failure occurrence time period of the failure data. And 7, determining a one-to-one corresponding relation between the residual value and the service life of the bearing according to the residual curve. Therefore, the time sequence processing model is trained based on the full-period data of the bearing sample, a more accurate model can be obtained, and a residual error curve between the time domain characteristics corresponding to the predicted value and the fault data can be obtained to accurately represent the residual error value between the predicted value and the fault data so as to prepare for subsequent use. As shown in fig. 5, fig. 5 is a residual error curve in the method for predicting the remaining life of the bearing shown in fig. 2. Thus, a time sequence processing model and a residual error curve can be obtained through training.
Wherein, the time domain characteristics in the 3 rd step are processed. Wherein the processing includes one or more of a plurality of time domain feature normalizations, a mean of residuals of the plurality of time domain features, and a sum of residuals of the plurality of time domain features. Training the time sequence processing model to be trained by the processed time domain characteristics further comprises normalizing the time domain characteristics to obtain normalized time domain characteristics, and training the time sequence processing model to be trained by the normalized time domain characteristics. Therefore, normalization processing can be performed, the data processing amount can be reduced, and the data processing efficiency can be improved. The processing includes selection and/or normalization.
And step 140, determining a residual value according to the estimated value and the time domain feature to be detected.
Fig. 6 is a schematic flowchart illustrating a specific process of step 130 and step 140 in the method for predicting the remaining life of the bearing shown in fig. 2.
In the embodiment shown in fig. 6, in combination with the first step in the step 130, further includes a step 231 of selecting any time-domain feature to be measured, which is used for characterizing the bearing degradation trend, from the time-domain features to be measured of the time-series data to be measured. And 232, performing normalization processing on any selected time domain feature to be measured to obtain the normalized time domain feature to be measured. And, the second step in step 130 may further include step 233, inputting the normalized time domain feature to be measured into the time sequence processing model to output an estimated value, wherein the time domain feature to be measured includes any time domain feature to be measured. The step 140 may further include a step 241 of determining a residual between the estimated value and any selected time domain feature to be measured as a residual value. Therefore, the time domain characteristic to be measured can be used as the bearing degradation index, the bearing degradation trend is better reflected, the calculated data volume is small, and the calculation efficiency of residual values is improved.
Fig. 7 is another detailed flowchart of steps 130 and 140 in the method for predicting the remaining life of a bearing shown in fig. 2.
In the embodiment shown in fig. 7, in combination with the first step in step 130, step 331 may further include selecting a plurality of time domain features to be measured, which characterize a bearing degradation trend, from the time domain features to be measured of the time series data to be measured. And 332, performing normalization processing on the selected multiple time domain features to be detected to obtain normalized time domain features to be detected. And, the second step in step 130 may further include step 333, inputting the normalized time domain feature to be measured into the time sequence processing model to output the estimated value. The time domain features to be measured may include a plurality of time domain features to be measured. Step 140 may further include step 341 of determining a plurality of residuals between the estimated value and the selected plurality of time-domain features to be measured. In step 342, an average of the plurality of residuals is determined as a residual value. Therefore, a plurality of time domain characteristics to be measured can be used as the degradation index of the bearing, so that the degradation trend of the bearing is reflected more comprehensively, and the calculation accuracy of the residual value is improved.
And 150, determining the bearing life corresponding to the residual value according to the one-to-one corresponding relation between the residual value and the bearing life, and taking the bearing life corresponding to the residual value as the residual life of the bearing to be predicted. The one-to-one correspondence between residual values and bearing lives in step 150 is a one-to-many correspondence between residual values in the residual curve and bearing lives, which is converted into a one-to-one correspondence between residual values and bearing lives. Illustratively, the time series data to be measured is, for example, data of the first 10 time points, and the time corresponding to the time series data to be measured is, for example, 10.51, so that the 11 th time point is predicted, and the next time corresponding to the time series data to be measured is, for example, 10.52. Specifically, the time point is related to the sampling frequency of the time series data to be measured.
The fault data comprises a fault occurrence starting moment and a fault occurrence time period of the fault data; the method further comprises the following steps: the one-to-one correspondence between residual values and bearing life is determined as follows: step 1, constructing a residual error curve according to residual error data of a bearing sample and the bearing service life of the bearing sample; and 2, fitting the residual error curve to obtain a fitted curve, wherein the residual error curve comprises a one-to-many corresponding relation between the residual error value and the service life of the bearing, and the fitted curve comprises a one-to-one corresponding relation between the residual error value and the service life of the bearing. Therefore, the bearing life which is uniquely corresponding to the residual error value can be obtained, and the residual life of the bearing can be conveniently calculated through the residual error value.
Further, the above 2 nd step may include using an exponential function y = a x And fitting a residual error curve corresponding to the fault occurrence time period to obtain a fitted curve, wherein the residual error curve comprises a one-to-many corresponding relation between residual error values and the service lives of the bearings, and the fitted curve comprises a one-to-one corresponding relation between the residual error values and the service lives of the bearings. The method is used because the degradation of the bearing is serious in the later life stage, the vibration is severe, the degradation development of the bearing fault is changed according to an exponential law instead of a linear law, and a residual error curve of a fault occurrence time period is continuously increasedThe exponential function of the embodiment is used for fitting the residual error curve corresponding to the fault occurrence time period, and further the residual life of the bearing to be predicted is obtained. The data amount for fitting the residual error curve corresponding to the fault occurrence time period is small, and the one-to-many corresponding relation between the residual error values in the residual error curve and the service life of the bearing can be converted into the one-to-one corresponding relation between the residual error values and the service life of the bearing, so that the residual life of the bearing can be conveniently calculated through the residual error values. In other embodiments of the present application, the residual curve of the fault occurrence time period may also be fitted by polynomial fitting, gaussian fitting, or other existing fitting manners, as long as the discrete residual values and the remaining bearing lives are converted into a one-to-one correspondence relationship.
The time point corresponding to the highest point of the fitting curve is T, the time point corresponding to the highest point of the fitting curve is also the end time point of the service life of the bearing, the true value of the RMS in the time domain feature is F, the estimated value of the RMS in the time domain feature is F, and the residual error value epsilon = F-F. And during the operation period of the bearing to be predicted, if the residual error value epsilon has a rising trend, substituting the residual error value epsilon into a fitting curve, and obtaining the bearing service life corresponding to the residual error value at the corresponding time point T as the residual service life RUL = T-T of the bearing to be predicted.
Fig. 8 is a schematic diagram illustrating an example of applying the peak-to-valley value in the time domain characteristic of the residual life prediction method for the bearing shown in fig. 2. FIG. 9 is a schematic diagram illustrating an example of the RMS application in the time domain characteristic of the method for predicting the remaining life of a bearing shown in FIG. 2. Fig. 10 is a schematic diagram illustrating an example of application of the pulse factor in the time domain characteristic of the residual life prediction method of the bearing shown in fig. 2. Fig. 11 is a diagram illustrating an example of application of kurtosis in a temporal characteristic of the method for predicting remaining life of a bearing shown in fig. 2. Fig. 12 is a schematic diagram illustrating an application example of the residual error sum curve and the fitting curve thereof in the method for predicting the residual life of the bearing shown in fig. 2. Fig. 13 is a schematic diagram illustrating an example of application of the residual life of the bearing in the method for predicting the residual life of the bearing shown in fig. 2.
Firstly, a plurality of time domain characteristics are extracted from the data of the full life cycle, and as can be seen from fig. 8 and 9, the peak-to-valley value and the RMS can well represent characteristic parameters of the degradation trend of the bearing, so that the numerical values of the two time domain characteristics are comprehensively considered, the numerical values of the two time domain characteristics are respectively input into a time domain processing model to output respective residual errors, and the average value of the residual errors of the numerical values of the two time domain characteristics is used as the degradation index of the bearing.
As shown in fig. 8 to 11, the bearing is in a healthy state from 5/1/2019 to 5/23/2019, the peak-to-valley value and RMS of the bearing start to rise after 5/23/2019, stabilize at about 0.015, and the fault starts to develop at this time, but the bearing still operates normally, the peak-to-valley value and RMS suddenly rise at 6/26/2019, and reach the highest value at 6/30/2019, and the fault is considered to end the life at this time, so that the fault occurrence time period from 17/2019 to 6/30/2019 can be considered as the fault occurrence time period, and the residual curve of the fault occurrence time period is fitted in the embodiment of the present application. The result of curve fitting of the sum of the residuals in the fault occurrence period is shown in fig. 12, and the residual life of the bearing is predicted, and the prediction result is shown in fig. 13.
Fig. 14 is a block diagram illustrating a residual life prediction apparatus of a bearing according to an embodiment of the present invention.
As shown in fig. 14, the device for predicting the remaining life of a bearing according to the embodiment of the present application includes the following modules:
an obtaining module 41, configured to obtain time sequence data to be predicted of a bearing;
the extraction module 42 is configured to extract a time domain feature of the time series data to be detected;
the first processing module 43 is configured to obtain an estimated value associated with the time domain feature to be detected by using a time sequence processing model according to the time domain feature to be detected, where the time sequence processing model is obtained by training full-period data of a bearing sample, the full-period data of the bearing sample includes health data of the bearing sample and fault data of the bearing sample, and the estimated value is obtained by using the time sequence processing model according to the time domain feature to be detected and the health data of the bearing sample;
a second processing module 44, configured to determine a residual value according to the estimated value and the time domain feature to be detected;
and the third processing module 45 is configured to determine, according to the one-to-one correspondence between the residual value and the bearing life, the bearing life corresponding to the residual value as the remaining life of the bearing to be predicted.
The implementation process of the functions and actions of each module in the above device is specifically described in the implementation process of the corresponding step in the above method, and is not described herein again.
Fig. 15 is a block diagram illustrating a structure of a residual life prediction system 50 of a bearing according to an embodiment of the present application.
As shown in fig. 15, the bearing remaining life prediction system 50 includes one or more processors 51 for implementing the bearing remaining life prediction method as described above.
In some embodiments, the bearing remaining life prediction system 50 may include a computer-readable storage medium 59, and the computer-readable storage medium 59 may store a program that may be invoked by the processor 51, and may include a non-volatile storage medium. In some embodiments, the bearing residual life prediction system 50 may include a memory 58 and an interface 57. In some embodiments, the bearing residual life prediction system 50 may also include other hardware depending on the actual application.
The computer-readable storage medium 59 of the embodiment of the present application has stored thereon a program for implementing the bearing remaining life prediction method as described above when the program is executed by the processor 51.
This application may take the form of a computer program product embodied on one or more computer-readable storage media 59 (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having program code embodied therein. Computer-readable storage media 59 include permanent and non-permanent, removable and non-removable media, and may implement information storage in any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer-readable storage media 59 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 technologies, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape storage or other magnetic storage devices, or any other non-transmission medium, may be used to store information that may be accessed by a computing device.
Embodiments of the present application also provide a computer program, which is stored in a computer-readable storage medium, such as the computer-readable storage medium in fig. 15, and when executed by a processor, causes the processor 31 to perform the method described above.
The above description is only a preferred embodiment of the present disclosure, and should not be taken as limiting the present disclosure, and any modifications, equivalents, improvements and the like made within the spirit and principle of the present disclosure should be included in the protection scope of the present disclosure.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, the phrases "comprising a" \8230; "defining an element do not exclude the presence of additional like elements in the process, method, article, or apparatus that comprises the element.
Claims (11)
1. A method for predicting the remaining life of a bearing, comprising:
acquiring time sequence data to be measured of a bearing to be predicted;
extracting the time domain characteristics to be detected of the time sequence data to be detected;
obtaining an estimated value associated with the time domain feature to be detected by using a time sequence processing model according to the time domain feature to be detected, wherein the time sequence processing model is obtained by training full-period data of a bearing sample, the full-period data of the bearing sample comprises health data of the bearing sample and fault data of the bearing sample, and the estimated value is obtained by using the time sequence processing model according to the time domain feature to be detected and the health data of the bearing sample;
determining a residual value according to the estimated value and the time domain feature to be detected;
and determining the bearing life corresponding to the residual value according to the one-to-one corresponding relation between the residual value and the bearing life, and taking the bearing life corresponding to the residual value as the residual life of the bearing to be predicted.
2. The method of predicting the residual life of a bearing according to claim 1, further comprising:
acquiring full-period data of a bearing sample;
extracting time domain characteristics of the full-period data;
processing the time domain features to obtain the processed time domain features, wherein the processed time domain features comprise time domain features corresponding to the health data and time domain features corresponding to the fault data;
training the time sequence processing model by using the processed time domain characteristics;
inputting the time domain characteristics corresponding to the health data into the time sequence processing model to obtain a predicted value associated with the time domain characteristics corresponding to the health data;
determining a residual error curve between the predicted value and the time domain characteristic corresponding to the fault data, wherein the residual error curve comprises a residual error value of the fault data; the fault data comprises a fault occurrence starting moment and a fault occurrence time period of the fault data;
and determining a one-to-one corresponding relation between the residual value and the service life of the bearing according to the residual curve.
3. The method of predicting the remaining life of a bearing according to claim 2,
the processing the time domain feature to obtain the processed time domain feature includes:
carrying out normalization processing on the time domain characteristics to obtain the normalized time domain characteristics;
the training of the time sequence processing model by using the processed time domain features and the obtaining of the predicted value associated with the time domain features to be tested comprise:
and training the time sequence processing model by using the normalized time domain characteristics.
4. The method of predicting remaining life of a bearing according to claim 1, further comprising:
acquiring a one-to-one corresponding relation between the residual value and the service life of the bearing by adopting the following method:
constructing a residual curve according to the residual data of the bearing sample and the bearing service life of the bearing sample;
and fitting the residual error curve to obtain a fitted curve, wherein the residual error curve comprises a one-to-many corresponding relation between the residual error value and the service life of the bearing, and the fitted curve comprises a one-to-one corresponding relation between the residual error value and the service life of the bearing.
5. The method for predicting the residual life of a bearing according to claim 4, wherein said fitting the residual curve to obtain a fitted curve comprises:
and fitting a residual error curve corresponding to the fault occurrence time period by using an exponential function to obtain a fitted curve, wherein the residual error curve comprises a one-to-many corresponding relation between residual error values and bearing service lives, and the fitted curve comprises a one-to-one corresponding relation between the residual error values and the bearing service lives.
6. A method of predicting remaining life of a bearing as claimed in any one of claims 1 to 5, wherein said time series processing model comprises a Bayes-LSTM model or a Bayes-GRU model.
7. The method for predicting the residual life of the bearing according to claim 1, wherein the obtaining the estimated value associated with the time domain feature to be measured by using a time sequence processing model according to the time domain feature to be measured comprises:
processing the time domain feature to be detected to obtain the processed time domain feature to be detected;
and inputting the processed time domain feature to be detected into the time sequence processing model so as to output the estimated value.
8. The method for predicting the residual life of the bearing according to claim 7, wherein the step of processing the time domain feature to be measured to obtain the processed time domain feature to be measured comprises: selecting any time domain feature to be detected representing the degradation trend of the bearing from the time domain features to be detected of the time sequence data to be detected; normalizing any selected time domain feature to be detected to obtain the normalized time domain feature to be detected; the inputting the processed time domain feature to be detected into the time sequence processing model to output the estimated value includes: inputting the normalized time domain feature to be detected into the time sequence processing model to output the estimated value; determining a residual value according to the estimated value and the time domain feature to be detected comprises: determining a residual error between the estimated value and any selected time domain feature to be detected as the residual error value;
and/or the presence of a gas in the gas,
the processing the time domain feature to be detected to obtain the normalized time domain feature to be detected includes: selecting a plurality of time domain characteristics to be detected representing the degradation trend of the bearing from the time domain characteristics to be detected of the time sequence data to be detected; normalizing the selected multiple time domain features to be detected to obtain normalized time domain features to be detected; the inputting the processed time domain feature to be detected into the time sequence processing model to output the estimated value includes: inputting the normalized time domain feature to be detected into the time sequence processing model to output the estimated value; determining a residual value according to the estimated value and the time domain feature to be detected comprises: determining a plurality of residual errors between the estimated value and the selected plurality of time domain features to be detected; determining an average value of the plurality of residuals as the residual value.
9. A bearing remaining life predicting device, comprising:
the system comprises an acquisition module, a prediction module and a prediction module, wherein the acquisition module is used for acquiring time sequence data to be predicted of a bearing to be predicted;
the extraction module is used for extracting the time domain characteristics to be detected of the time sequence data to be detected;
the time domain characteristic measuring device comprises a first processing module, a time sequence processing module and a second processing module, wherein the first processing module is used for obtaining an estimated value related to the time domain characteristic to be measured by utilizing the time sequence processing model according to the time domain characteristic to be measured, the time sequence processing model is obtained by training full-period data of a bearing sample, the full-period data of the bearing sample comprises health data of the bearing sample and fault data of the bearing sample, and the estimated value is obtained by utilizing the time sequence processing model according to the time domain characteristic to be measured and the health data of the bearing sample;
the second processing module is used for determining a residual value according to the estimated value and the time domain feature to be detected;
and the third processing module is used for determining the bearing life corresponding to the residual value according to the one-to-one corresponding relation between the residual value and the bearing life, and the bearing life is used as the residual life of the bearing to be predicted.
10. A bearing residual life prediction system comprising one or more processors configured to implement the bearing residual life prediction method according to any one of claims 1 to 8.
11. A computer-readable storage medium, having stored thereon a program which, when executed by a processor, implements a method of predicting remaining life of a bearing according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211098106.0A CN115438443A (en) | 2022-09-08 | 2022-09-08 | Bearing residual life prediction method, device and system and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211098106.0A CN115438443A (en) | 2022-09-08 | 2022-09-08 | Bearing residual life prediction method, device and system and readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115438443A true CN115438443A (en) | 2022-12-06 |
Family
ID=84246902
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211098106.0A Pending CN115438443A (en) | 2022-09-08 | 2022-09-08 | Bearing residual life prediction method, device and system and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115438443A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116990021A (en) * | 2023-09-22 | 2023-11-03 | 万向钱潮股份公司 | Fatigue life assessment method and device for hub bearing |
-
2022
- 2022-09-08 CN CN202211098106.0A patent/CN115438443A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116990021A (en) * | 2023-09-22 | 2023-11-03 | 万向钱潮股份公司 | Fatigue life assessment method and device for hub bearing |
| CN116990021B (en) * | 2023-09-22 | 2024-01-02 | 万向钱潮股份公司 | Fatigue life assessment method and device for hub bearing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Oh et al. | A novel method and its field tests for monitoring and diagnosing blade health for wind turbines | |
| US20190226943A1 (en) | Facility condition monitoring device and facility condition monitoring method | |
| US20120209539A1 (en) | Turbine fault analysis | |
| Kusnick et al. | Wind turbine rotor imbalance detection using nacelle and blade measurements | |
| CN115438443A (en) | Bearing residual life prediction method, device and system and readable storage medium | |
| CN114423940A (en) | Method for estimating remaining useful life of a component of an operational wind turbine | |
| Almeida et al. | Experimental analysis of dynamic interaction between a centrifugal compressor and its casing | |
| Xing et al. | Detection of magnitude and position of rotor aerodynamic imbalance of wind turbines using Convolutional Neural Network | |
| KR102226971B1 (en) | Method for fault diagnosis based on multiple variables and apparatus using the method | |
| Myrent et al. | Aerodynamic sensitivity analysis of rotor imbalance and shear web disbond detection strategies for offshore structural health prognostics management of wind turbine blades | |
| Maturkanič et al. | Construction of the signal profile for use in blade tip-timing analysis | |
| US20240287968A1 (en) | Method and device for evaluating service life of pitch bearing of wind turbine | |
| Zhang et al. | Probability warning for wind turbine gearbox incipient faults based on SCADA data | |
| Myrent et al. | Wind turbine blade shear web disbond detection using rotor blade operational sensing and data analysis | |
| Mehlan et al. | Estimation of wind turbine gearbox loads for online fatigue monitoring using inverse methods | |
| Ahilan et al. | A Self-Improved Optimizer-Based CNN for Wind Turbine Fault Detection | |
| Nejad | Modelling and analysis of drivetrains in offshore wind turbines | |
| Daems et al. | Gaining insights in loading events for wind turbine drivetrain prognostics | |
| Shahriar et al. | Fault detection of wind turbine drivetrain utilizing power-speed characteristics | |
| da Rosa et al. | Wind Turbine Blade Mass Imbalance Detection Using Artificial Intelligence | |
| Myrent et al. | Pitch error and shear web disbond detection on wind turbine blades for offshore structural health and prognostics management | |
| Daems et al. | Farm-wide dynamic event classification as load input for wind turbine drivetrain lifetime prognosis | |
| CN113669214B (en) | Method, system and storage medium for detecting running state of planetary stage of wind power gear box | |
| CN116256170A (en) | Bearing state detection method, apparatus and computer readable storage medium | |
| Wang et al. | The absolute deviation rank diagnostic approach to gear tooth composite fault |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |