CN116625688A - Rolling bearing health monitoring method based on multilayer noise reduction and self-encoder - Google Patents
Rolling bearing health monitoring method based on multilayer noise reduction and self-encoder Download PDFInfo
- Publication number
- CN116625688A CN116625688A CN202310592153.9A CN202310592153A CN116625688A CN 116625688 A CN116625688 A CN 116625688A CN 202310592153 A CN202310592153 A CN 202310592153A CN 116625688 A CN116625688 A CN 116625688A
- Authority
- CN
- China
- Prior art keywords
- noise reduction
- rolling bearing
- encoder
- health
- self
- 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
- 230000009467 reduction Effects 0.000 title claims abstract description 53
- 230000036541 health Effects 0.000 title claims abstract description 44
- 238000005096 rolling process Methods 0.000 title claims abstract description 38
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000012544 monitoring process Methods 0.000 title claims abstract description 12
- 238000012512 characterization method Methods 0.000 claims abstract description 18
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 16
- 238000012216 screening Methods 0.000 claims abstract description 10
- 238000011156 evaluation Methods 0.000 claims description 10
- 238000010606 normalization Methods 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 6
- 230000003862 health status Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 description 6
- 238000012549 training Methods 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 4
- 230000007613 environmental effect Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000006731 degradation reaction Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 238000000638 solvent extraction Methods 0.000 description 3
- 238000003745 diagnosis Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 229910052956 cinnabar Inorganic materials 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000036039 immunity Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
- G01M13/045—Acoustic or vibration analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
- G06F18/15—Statistical pre-processing, e.g. techniques for normalisation or restoring missing data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/211—Selection of the most significant subset of features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- 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
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2123/00—Data types
- G06F2123/02—Data types in the time domain, e.g. time-series data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Acoustics & Sound (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The application discloses a rolling bearing health monitoring method based on multilayer noise reduction and a self-encoder, which comprises the following steps: performing LMD decomposition, wavelet threshold noise reduction and SVD decomposition treatment on an original vibration signal of the rolling bearing, and reconstructing the treated signal to obtain a noise reduction signal; the characteristic features of the noise reduction signals are obtained by using the characteristics of encoding and decoding of a self-encoder, wherein both the encoder and the decoder are realized by an LSTM network and a full connection layer; selecting effective characteristics in the characterization characteristics, comprehensively evaluating the robustness, the predictability and the relativity of the effective characteristics, and screening out the characteristics with the best comprehensive performance as health indexes; inputting the obtained health index into an LSTM network, and obtaining a predicted value capable of reflecting the health state of the bearing through the LSTM network. The method can obtain the health state of the rolling bearing and improve the accuracy of prediction.
Description
Technical Field
The application relates to the technical field of bearing fault diagnosis, in particular to a rolling bearing health monitoring method based on multilayer noise reduction and a self-encoder.
Background
Rolling bearings are widely used in mechanical equipment, and in actual operation, due to different working conditions and environmental influences, collected vibration signals often contain high-intensity noise signals, which can have great adverse effects on the evaluation of the health state of the rolling bearings by extracting characterization features from the vibration signals. Therefore, research on vibration signal feature extraction and prediction under a strong noise background is of great significance to effective operation of the rolling bearing.
The noise reduction processing is carried out on the original vibration signal, so that a large amount of environmental noise and redundant information in the data can be eliminated, the data quality is improved, the representation capability of health indexes is further improved, and the prediction accuracy is improved. The good denoising effect is greatly helpful for signal analysis, and the model performance is improved to a great extent. For example, the method of combining inherent time scale analysis with minimum entropy deconvolution with multi-point most adjustment is utilized by cinnabar and the like to perform noise reduction treatment on an original vibration signal, and a better effect is obtained in rolling bearing fault diagnosis. The method is used for extracting transient impact signals by adopting a self-adaptive sparse decomposition method so as to eliminate the influence of high-amplitude interference impact and background noise in vibration signals. The method for reducing the noise of the vibration data by combining the VMD with the adaptive rolling bearing with the minimum entropy deconvolution of multiple points is proposed by Luo et al, and better characterization features are extracted from the vibration signals. However, the accuracy of the prediction is generally low in the prior art, and the health state of the rolling bearing cannot be accurately estimated.
Disclosure of Invention
The application aims to solve the technical problem of providing a rolling bearing health monitoring method with high prediction accuracy based on multi-layer noise reduction and a self-encoder.
In order to solve the technical problems, the application adopts the following technical scheme: a rolling bearing health monitoring method based on multilayer noise reduction and self-encoder is characterized by comprising the following steps:
data noise reduction: performing LMD decomposition, wavelet threshold noise reduction and SVD decomposition treatment on an original vibration signal of the rolling bearing, and reconstructing the treated signal to obtain a noise reduction signal;
normalization: converting the value of the noise reduction signal into a range of [0,1] using a max-min normalization;
feature extraction: the characteristic features of the noise reduction signals are obtained by using the characteristics of encoding and decoding of a self-encoder, wherein both the encoder and the decoder are realized by an LSTM network and a full connection layer;
feature screening: selecting effective characteristics in the characterization characteristics, comprehensively evaluating the robustness, the predictability and the relativity of the effective characteristics, and screening out the characteristics with the best comprehensive performance as health indexes;
health status assessment: inputting the obtained health index into an LSTM network, and obtaining a predicted value capable of reflecting the health state of the bearing through the LSTM network.
The further technical scheme is that the data denoising step specifically comprises the following steps:
carrying out LMD (least mean squares) decomposition on the original data of the input vibration signals to obtain a plurality of PF components;
and calculating the correlation coefficient of each PF component and the original data, screening the PF components according to the correlation coefficient, and dividing the PF components into effective PF components and other PF components according to the size of the correlation coefficient, wherein the calculation formula of the correlation coefficient is as follows:
wherein: c (C) i Represents the i-th correlation coefficient, n represents the length of the input data and the PF component, x (j) represents the j-th value of the input data,representing the average value of the input data, PF i (j) Represents the j-th value in the i-th PF component,>representing the average value of the ith PF component;
the effective PF components are subjected to wavelet threshold noise reduction respectively, and the noise reduction structure is reconstructed;
SVD decomposition is carried out on the result processed by the steps to obtain a reconstruction signal;
and adding and fusing the reconstructed signal and other PF components to obtain final noise reduction data.
The further technical proposal is that the acquisition of the health index comprises the following steps:
for the feature sequence f= [ F (T1), F (T2), F (T3) & gt.f (tN) ] and the time sequence t= [ T1, T2, t3.. & gt.tn ], where N is the total length of the sequence, the calculation formula for the correlation Corr (F, T), the predictive Pre (F), the robustness Rob (F) is:
the extracted multidimensional features are evaluated by the three indexes, a certain weight is given to each index, a total evaluation score of the features in each dimension is obtained as shown in a formula (5), and the highest feature is selected as a health index of the rolling bearing:
Score＝0.35×Corr(F，T)+0.4×Pre(F)+0.25×Rob(F) (5)；
wherein Corr (F, T) represents the correlation of the feature sequence with the time sequence, and Pre (F) represents the predictability of the feature sequence; rob (F) represents the robustness of the feature sequence.
The beneficial effects of adopting above-mentioned technical scheme to produce lie in: according to the method, firstly, the original data of the vibration signal is preprocessed by using a LMD, WTD, SVD-based multi-layer joint noise reduction method, the characterization features of the rolling bearing are extracted and screened from the processed data by using an LSTM-based self-encoder network to serve as health indexes, and then the health indexes are input into a prediction model to obtain the health state of the rolling bearing. The method is validated using the public data set and validated by comparison with the original vibration signal as input and other models.
Drawings
The application will be described in further detail with reference to the drawings and the detailed description.
FIG. 1 is a flow chart of a method according to an embodiment of the application;
FIG. 2 is a schematic block diagram of multi-layer joint noise reduction in a method according to an embodiment of the application;
FIG. 3 is a diagram of a feature extraction process in a method according to an embodiment of the application;
FIG. 4 is a comparison of the data noise reduction before and after the data noise reduction in the embodiment of the application;
FIG. 5 is a 37 th dimension feature diagram in an embodiment of the application;
FIG. 6 is a graph of training results for Bearing1_6 in an embodiment of the application;
FIG. 7 is a graph of LOSS decline in an embodiment of the application;
FIG. 8 is a predicted outcome of a method according to an embodiment of the application;
FIG. 9 shows a second prediction result of the method according to the embodiment of the application.
Detailed Description
The following description of the embodiments of the present application will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are only some, but not all embodiments of the application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application, but the present application may be practiced in other ways other than those described herein, and persons skilled in the art will readily appreciate that the present application is not limited to the specific embodiments disclosed below.
In general, as shown in fig. 1, the embodiment of the application discloses a rolling bearing health monitoring method based on multi-layer noise reduction and self-encoder, which is characterized in that in order to reduce the influence of external environment in rolling bearing health state monitoring, a multi-layer combined noise reduction module based on LMD, wavelet threshold noise reduction and SVD is built to reduce noise information of vibration signals and make full use of effective information. In order to fully utilize the advantages of the LSTM network and the self-encoder for processing the time sequence and extracting the data characterization features, a self-encoder module based on the LSTM network is built for extracting the characterization features of the rolling bearing. And then comprehensively evaluating the extracted characterization features, using the feature with the highest evaluation index as a health index of the rolling bearing, and inputting the health index into the LSTM model to predict the performance degradation degree of the rolling bearing. The method specifically comprises the following steps:
1) Data noise reduction: and carrying out data denoising on the original vibration signal by LMD decomposition, wavelet threshold denoising and SVD decomposition, and reconstructing the denoised signal to obtain a denoising signal.
2) Normalization: the value of the noise reduction signal is converted into the range of [0,1] using max-min normalization for which prediction accuracy can be improved.
3) Feature extraction: the characterization features of the noise reduction signal are obtained using the characteristics of the self-encoder encoding, decoding. Both the encoder and the decoder are implemented by an LSTM network and a fully connected layer, resulting in a hidden layer feature of 100 dimensions.
4) Feature screening: and selecting effective characteristics in the characterization characteristics, comprehensively evaluating the robustness, the predictability and the relativity of the effective characteristics, and screening out the characteristics with the best comprehensive performance as health indexes.
5) Health status assessment: inputting the health index obtained in the step 4) into an LSTM network, and obtaining a predicted value capable of reflecting the health state of the bearing through the LSTM network
Multi-layer joint noise reduction
Due to the influence of environmental noise, the vibration signal has the characteristics of nonlinearity, non-stability and multi-source vibration coupling. Eliminating the influence of environmental noise has important significance for analyzing vibration signals, extracting effective characteristics and improving prediction accuracy.
The LMD has excellent non-stationary signal processing capability. The problem of over-envelope and under-envelope of an empirical mode decomposition algorithm is effectively solved while the non-stationary signal is decomposed into a plurality of PF components. WTD may extract noise from a set of data or signals and translate to finding an optimal threshold in the data to suppress noise interference. The SVD method can effectively remove interference signals such as noise, clutter and the like, simultaneously retain useful information in the signals, and improve the accuracy and reliability of signal processing.
In combination with the excellent characteristics of LMD, WTD and SVD in extracting signal period components and filtering random noise, a multi-layer joint noise reduction scheme is proposed, and the noise reduction structure is shown in fig. 2:
the specific noise reduction flow is as follows:
1) Carrying out LMD (least mean squares) decomposition on the original data of the input vibration signals to obtain a plurality of PF components;
2) And calculating the correlation coefficient of each PF component and the original data, screening the PF components according to the correlation coefficient, and dividing the PF components into effective PF components and other PF components according to the size of the correlation coefficient, wherein the calculation formula of the correlation coefficient is as follows:
wherein C is i Represents the i-th correlation coefficient, n represents the length of the input data and the PF component, x (j) represents the j-th value of the input data,representing the average value of the input data, PF i (j) Represents the j-th value in the i-th PF component,>the average value of the ith PF component is represented.
3) The effective PF components are subjected to wavelet threshold noise reduction respectively, and the noise reduction structure is reconstructed;
4) SVD (singular value decomposition) is carried out on the result processed in the step 3) to obtain a reconstruction signal;
5) And adding and fusing the reconstructed signal and other PF components to obtain final noise reduction data.
Feature extraction and evaluation
The LSTM network can well mine the relation information among the time sequence data at different moments through a gating mechanism thereof, has good time sequence processing capability, and can mine effective time sequence characteristics. The self-encoder forces the model to learn low-dimensional characteristics to characterize the original data through the encoding and decoding processes. Combining LSTM and self-coding, a self-coder based on LSTM network is constructed for extracting the time sequence characteristics of bearing vibration signals. The time dependence of the vibration signals is better processed, the capacity of the encoder for extracting the characterization vibration signals is improved, and the problems of gradient explosion and gradient disappearance are solved.
As shown in fig. 3, the self-encoder is composed of an LSTM network and a fully connected network. The noise-reduced vibration signal is subjected to normalization processing, the noise-reduced signal value is converted into the range of [0,1] by using maximum-minimum normalization to improve training precision, then the noise-reduced signal value is input into a self-encoder, the characterization features of a time sequence are extracted through an LSTM network, the characterization features are synthesized through a full-connection layer to obtain hidden layer features capable of characterizing input data, then the hidden layer features are input into a decoder, and data similar to the input signal is obtained through reverse slave operation of the full-connection network and the LSTM network. And extracting hidden layer characteristics as characterization characteristics of the rolling bearing in the process of model training.
The degradation process of the rolling bearing is a time sequence process, and the proper characteristic parameters can be used as health indexes to well describe the degradation process of the rolling bearing. Correlation, predictability and robustness are three key indicators for evaluating the magnitude of the characteristic parameter characterization capability. The correlation is used for evaluating the degree of correlation between the characteristic parameter and the time sequence vibration sequence; monotonicity is used to describe the trend of a feature sequence continuously increasing or decreasing; robustness is used to reflect the anti-interference or noise immunity of the characteristic signal.
For the feature sequence f= [ F (T1), F (T2), F (T3) & gt.f (tN) ] and the time sequence t= [ T1, T2, t3.. & gt.tn ], where N is the total length of the sequence, the calculation formula for the correlation Corr (F, T), the predictive Pre (F), the robustness Rob (F) is:
the extracted multidimensional features are evaluated by the three indexes, a certain weight is given to each index, a total evaluation score of the features in each dimension is obtained as shown in a formula (5), and the highest feature is selected as a health index of the rolling bearing:
Score＝0.35×Corr(F，T)+0.4×Pre(F)+0.25×Rob(F) (5)；
wherein Corr (F, T) represents the correlation of the feature sequence with the time sequence, and Pre (F) represents the predictability of the feature sequence; rob (F) represents the robustness of the feature sequence.
Experimental verification results and analysis
Data set partitioning:
the test in this chapter uses the PHM2014 bearing data set obtained from the PRONOSTIA platform as the data set of working condition 1. See section 3.3.1 for details. There are 7 rolling bearing life-time data from run to failure under condition 1. In the experiment, the life-span data of the first 6 rolling bearings are used as training set, and the 7 th bearing data are used as test set. The data set partitioning is shown in table 1.
Table 1 data set partitioning case
Experimental verification results and analysis
Firstly, the original vibration signal is preprocessed through the multi-layer joint noise reduction module. The noise is obviously suppressed before and after the noise reduction of part of the data, as shown in fig. 4. The noise-reduced data is smoother than the original data, and the excessive amplitude due to noise is also reduced to some extent.
And then extracting the characterization features of the vibration data through the LSTM-based self-encoder network provided by the method and the encoding and decoding processes. In the experiment, 100-dimensional features were extracted altogether, with 50-dimensional effective features altogether. And evaluating the characteristics of each dimension by using the evaluation index. The feature that gives the highest composite score is the 37 th dimension feature, as shown in fig. 5. The 37 th dimension characteristic is input into the prediction model as a health index for monitoring the health state of the rolling bearing.
When the LSTM network is used as a prediction model of the health state of the bearing, the training round number epochs is 50, the optimizer dimension Adam is set as a mean square error loss function, and the training results are shown in fig. 6 and 7.
The method described in the application was compared with a method without noise reduction module, the prediction results are shown in fig. 7 and 8, and the price index is shown in table 2, using RMSE and MAE as evaluation indexes of the model, in order to verify the effectiveness and superiority of the experiment. It is known that the prediction result obtained after the noise reduction treatment is performed on the vibration signal of the rolling bearing is superior to the prediction result of the original vibration signal in both evaluation indexes.
Table 2 evaluation index results
In summary, the method of the application firstly uses a LMD, WTD, SVD-based multi-layer joint noise reduction method to preprocess the original data of the vibration signal, uses an LSTM-based self-encoder network to extract and screen the characterization characteristics of the rolling bearing from the processed data as health indexes, and then inputs the health indexes into a prediction model to obtain the health state of the rolling bearing. The method is validated using the public data set and validated by comparison with the original vibration signal as input and other models.
Claims (3)
1. A rolling bearing health monitoring method based on multilayer noise reduction and self-encoder is characterized by comprising the following steps:
data noise reduction: performing LMD decomposition, wavelet threshold noise reduction and SVD decomposition treatment on an original vibration signal of the rolling bearing, and reconstructing the treated signal to obtain a noise reduction signal;
normalization: converting the value of the noise reduction signal into a range of [0,1] using a max-min normalization;
feature extraction: the characteristic features of the noise reduction signals are obtained by using the characteristics of encoding and decoding of a self-encoder, wherein both the encoder and the decoder are realized by an LSTM network and a full connection layer;
feature screening: selecting effective characteristics in the characterization characteristics, comprehensively evaluating the robustness, the predictability and the relativity of the effective characteristics, and screening out the characteristics with the best comprehensive performance as health indexes;
health status assessment: inputting the obtained health index into an LSTM network, and obtaining a predicted value capable of reflecting the health state of the bearing through the LSTM network.
2. The rolling bearing health monitoring method based on multi-layer noise reduction and self-encoder as claimed in claim 1, characterized in that the step of data noise reduction comprises the steps of:
carrying out LMD (least mean squares) decomposition on the original data of the input vibration signals to obtain a plurality of PF components;
and calculating the correlation coefficient of each PF component and the original data, screening the PF components according to the correlation coefficient, and dividing the PF components into effective PF components and other PF components according to the size of the correlation coefficient, wherein the calculation formula of the correlation coefficient is as follows:
wherein: c (C) i Represents the i-th correlation coefficient, n represents the length of the input data and the PF component, x (j) represents the j-th value of the input data,representing the average value of the input data, PF i (j) Represents the j-th value in the i-th PF component,>representing the average value of the ith PF component;
the effective PF components are subjected to wavelet threshold noise reduction respectively, and the noise reduction structure is reconstructed;
SVD decomposition is carried out on the result processed by the steps to obtain a reconstruction signal;
and adding and fusing the reconstructed signal and other PF components to obtain final noise reduction data.
3. The rolling bearing health monitoring method based on multi-layer noise reduction and self-encoder according to claim 1, wherein the obtaining of the health index comprises the steps of:
for the feature sequence f= [ F (T1), F (T2), F (T3) & gt.f (tN) ] and the time sequence t= [ T1, T2, t3.. & gt.tn ], where N is the total length of the sequence, the calculation formula for the correlation Corr (F, T), the predictive Pre (F), the robustness Rob (F) is:
the extracted multidimensional features are evaluated by the three indexes, a certain weight is given to each index, a total evaluation score of the features in each dimension is obtained as shown in a formula (5), and the highest feature is selected as a health index of the rolling bearing:
Score＝0.35×Corr(F，T)+0.4×Pre(F)+0.25×Rob(F) (5)；
wherein Corr (F, T) represents the correlation of the feature sequence with the time sequence, and Pre (F) represents the predictability of the feature sequence; rob (F) represents the robustness of the feature sequence.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310592153.9A CN116625688A (en) | 2023-05-24 | 2023-05-24 | Rolling bearing health monitoring method based on multilayer noise reduction and self-encoder |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310592153.9A CN116625688A (en) | 2023-05-24 | 2023-05-24 | Rolling bearing health monitoring method based on multilayer noise reduction and self-encoder |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116625688A true CN116625688A (en) | 2023-08-22 |
Family
ID=87620845
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310592153.9A Pending CN116625688A (en) | 2023-05-24 | 2023-05-24 | Rolling bearing health monitoring method based on multilayer noise reduction and self-encoder |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116625688A (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102829974A (en) * | 2012-08-07 | 2012-12-19 | 北京交通大学 | LMD (local mean decomposition) and PCA (principal component analysis) based rolling bearing state identification method |
| CN109000921A (en) * | 2017-06-07 | 2018-12-14 | 中国大唐集团科技工程有限公司 | A kind of diagnostic method of wind generator set main shaft failure |
| CN109975013A (en) * | 2019-04-30 | 2019-07-05 | 南京工程学院 | Gearbox of wind turbine fault signature extracting method based on IVMD-SE |
| CN110146292A (en) * | 2019-06-04 | 2019-08-20 | 昆明理工大学 | A kind of rolling bearing fault testing method that the overall local mean value based on the reconstruct of integrated noise is decomposed |
| CN112069621A (en) * | 2020-09-08 | 2020-12-11 | 西安电子科技大学 | Method for predicting residual service life of rolling bearing based on linear reliability index |
| CN114841208A (en) * | 2022-05-14 | 2022-08-02 | 哈尔滨理工大学 | Rolling bearing performance decline prediction method and device based on SAE and TCN-Attention model |
| WO2023014007A1 (en) * | 2021-08-06 | 2023-02-09 | 주식회사 제이엘케이바이오 | Device and method for extracting compound information |
| CN116086807A (en) * | 2023-01-31 | 2023-05-09 | 北华大学 | Fault judging method for rolling bearing and related product |
-
2023
- 2023-05-24 CN CN202310592153.9A patent/CN116625688A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102829974A (en) * | 2012-08-07 | 2012-12-19 | 北京交通大学 | LMD (local mean decomposition) and PCA (principal component analysis) based rolling bearing state identification method |
| CN109000921A (en) * | 2017-06-07 | 2018-12-14 | 中国大唐集团科技工程有限公司 | A kind of diagnostic method of wind generator set main shaft failure |
| CN109975013A (en) * | 2019-04-30 | 2019-07-05 | 南京工程学院 | Gearbox of wind turbine fault signature extracting method based on IVMD-SE |
| CN110146292A (en) * | 2019-06-04 | 2019-08-20 | 昆明理工大学 | A kind of rolling bearing fault testing method that the overall local mean value based on the reconstruct of integrated noise is decomposed |
| CN112069621A (en) * | 2020-09-08 | 2020-12-11 | 西安电子科技大学 | Method for predicting residual service life of rolling bearing based on linear reliability index |
| WO2023014007A1 (en) * | 2021-08-06 | 2023-02-09 | 주식회사 제이엘케이바이오 | Device and method for extracting compound information |
| CN114841208A (en) * | 2022-05-14 | 2022-08-02 | 哈尔滨理工大学 | Rolling bearing performance decline prediction method and device based on SAE and TCN-Attention model |
| CN116086807A (en) * | 2023-01-31 | 2023-05-09 | 北华大学 | Fault judging method for rolling bearing and related product |
Non-Patent Citations (3)
| Title |
|---|
| 吕靖香;余建波;: "基于多层混合滤噪的轴承早期弱故障特征提取方法", 《振动与冲击》, vol. 37, no. 8, pages 22 - 27 * |
| 谢霖铨;梁博群;: "基于降噪自编码的推荐算法", 计算机与现代化, no. 02 * |
| 赵伟杰;杨乐乐;郝旺身;郝伟;: "奇异值分解与LMD结合的滚动轴承故障诊断研究", 机械设计与制造, no. 05 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110879254B (en) | Steel rail crack acoustic emission signal detection method based on improved least square generation type countermeasure network | |
| CN110550518A (en) | Elevator operation abnormity detection method based on sparse denoising self-coding | |
| CN105258940B (en) | The standardization m ultiwavelet that mechanical breakdown is quantitatively extracted and multi-wavelet packets transform method | |
| CN109443752B (en) | Gear vibration signal noise reduction and fault diagnosis method based on VMD | |
| Kabir et al. | Image compression using lifting based wavelet transform coupled with SPIHT algorithm | |
| Shan et al. | A motor bearing fault voiceprint recognition method based on Mel-CNN model | |
| CN115062733B (en) | Transformer voiceprint fault diagnosis method based on empirical mode decomposition and butterfly algorithm | |
| Zhang et al. | Bearing fault diagnosis under various operation conditions using synchrosqueezing transform and improved two-dimensional convolutional neural network | |
| Chao et al. | Improving accuracy of cavitation severity recognition in axial piston pumps by denoising time–frequency images | |
| CN116578940A (en) | Bearing fault new type identification and diagnosis method based on mixed depth self-coding | |
| Sun et al. | Data-driven fault diagnosis method based on second-order time-reassigned multisynchrosqueezing transform and evenly mini-batch training | |
| Zhang et al. | An adaptive graph morlet wavelet transform for railway wayside acoustic detection | |
| Li et al. | Incipient detection of bearing fault using impulse feature enhanced weighted sparse representation | |
| Yao et al. | RUL prediction method for rolling bearing using convolutional denoising autoencoder and bidirectional LSTM | |
| Yu et al. | Sparse time–frequency representation for the transient signal based on low-rank and sparse decomposition | |
| CN116625688A (en) | Rolling bearing health monitoring method based on multilayer noise reduction and self-encoder | |
| CN112363050A (en) | SF6 circuit breaker arc contact state evaluation method based on dynamic contact resistance signal | |
| LU502454B1 (en) | A Bearing Fault Identification Method and System Based on EEMD Sparse Decomposition | |
| CN116484176A (en) | Bearing fault diagnosis method, system and storage medium based on ultra-wavelet | |
| CN113409213B (en) | Method and system for enhancing noise reduction of time-frequency diagram of fault signal of plunger pump | |
| CN115733673A (en) | Data anomaly detection method based on multi-scale residual classifier | |
| Li et al. | Discrimination and prediction of tool wear state based on gray theory | |
| CN110688981B (en) | Modal aliasing elimination method for denoising vibration signal | |
| Jia et al. | Elevator running fault monitoring method based on vibration signal | |
| Lingli et al. | A novel MPE‐LPP‐ELM recognition method for the fault diagnosis of spiral bevel gears |
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 |