CN112990524A - Residual error correction-based residual life prediction method for rolling bearing - Google Patents
Residual error correction-based residual life prediction method for rolling bearing Download PDFInfo
- Publication number
- CN112990524A CN112990524A CN201911290741.7A CN201911290741A CN112990524A CN 112990524 A CN112990524 A CN 112990524A CN 201911290741 A CN201911290741 A CN 201911290741A CN 112990524 A CN112990524 A CN 112990524A
- Authority
- CN
- China
- Prior art keywords
- residual
- rolling bearing
- model
- life
- error correction
- 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
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
- 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
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Acoustics & Sound (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses a method for predicting the residual life of a rolling bearing by using a long and short memory network (LSTM) and a residual error correction mode, belongs to the technical field of numerical control, and particularly relates to a residual life prediction method of a rolling bearing based on residual error correction. The invention comprises the following steps: collecting vibration acceleration signals of the whole life cycle of the rolling bearing, and training a residual life prediction model by using the signals; processing the vibration acceleration signal in a sliding window mode to obtain a training set, and training an LSTM network model by using the training set; training a BP neural network model for predicting residual errors; constructing a residual error correction model, and adding the result of the LSTM network model and the result of the BP neural network model to obtain a prediction result of a final model; and inputting a signal for predicting the residual life of the rolling bearing into the residual error correction model to obtain the residual life at the current moment. The invention improves the accuracy of the residual life prediction.
Description
Technical Field
The invention discloses a method for predicting the residual life of a rolling bearing by using a long and short memory network (LSTM) and a residual error correction mode, belongs to the technical field of numerical control, and particularly relates to a residual life prediction method of a rolling bearing based on residual error correction.
Background
The rolling bearing is one of the widely used parts on the numerical control machine tool. According to statistics, 30% of faults of the rotating machinery are caused by bearing faults, and the health state of the bearing is closely related to whether the mechanical energy can normally run, so that a residual life prediction model of the rolling bearing is established, the degradation degree of the rolling bearing is evaluated in real time, and the bearing which cannot meet the working requirement is replaced in time according to the use requirement of the machinery, so that the important significance for ensuring the healthy and stable running of the machinery is realized.
The prediction of the residual life of the rolling bearing is researched by numerous scholars at home and abroad, and numerous methods are proposed. The traditional thinking is that a performance degradation index is established; and establishing a prediction model for prediction.
The performance degradation indicator is constructed by extracting single statistical features such as root mean square and kurtosis from the original signal. The performance degradation index is manually constructed by using the single statistical characteristic, a large amount of professional knowledge and experience are needed, and the subsequent prediction accuracy is difficult to guarantee by the single statistical characteristic.
The prediction model is constructed in a traditional machine learning regression prediction model, such as SVR or by using a hidden Markov model, but the methods need manual feature extraction, and the model is simple and has low prediction accuracy.
Disclosure of Invention
The invention provides a novel method for predicting residual life based on LSTM and residual correction, aiming at overcoming the defects that performance degradation indexes are difficult to construct and characteristics are difficult to extract in the traditional mode and the life prediction effect of a traditional prediction model is poor.
The technical scheme adopted by the invention for realizing the purpose is as follows:
the residual error correction-based residual life prediction method for the rolling bearing comprises the following steps of:
1) collecting vibration acceleration signals of the whole life cycle of a rolling bearing;
2) processing the vibration acceleration signal in a sliding window mode to obtain a training set, and training an LSTM network model by using the training set;
3) training a BP neural network model for predicting residual errors;
4) constructing a residual error correction model, and adding the result of the LSTM network model and the result of the BP neural network model to obtain a prediction result of a final model;
5) and inputting a signal for predicting the residual life of the rolling bearing into the residual error correction model to obtain the residual life at the current moment.
The input of the LSTM network model is the characteristics of the rolling bearing at the time t obtained in a sliding window mode; and outputting the service life of the rolling bearing predicted by the LSTM network model at the time t.
The input of the BP neural network model is a vibration acceleration signal at the time t, and the output is a residual error predicted by the BP neural network model at the time t.
The algorithm of the residual error is as follows:
Q＝S-L
and Q is a residual error, S is the real residual life of the rolling bearing corresponding to the vibration acceleration signal of the rolling bearing measured at the time t, and L is the residual life of the rolling bearing predicted by the LSTM model at the time t.
The residual error correction model is as follows:
R＝R1(x1)+R2(x2)
wherein, R1(x1) is used as L to be the life of the rolling bearing predicted by the LSTM model, R2(x2) is used as Q to be the residual error predicted by the BP neural network, and R is the residual life of the rolling bearing.
The invention has the following beneficial effects and advantages:
the method does not need a complex statistical learning process or special pretreatment of the bearing information, and improves the accuracy of residual life prediction.
Drawings
FIG. 1 is a flow chart of a method of the present invention;
FIG. 2 is a flow chart of training an LSTM network model;
FIG. 3 is a flowchart of a BP neural network for training prediction residuals;
fig. 4 is a schematic view of the sliding process.
Detailed Description
The scheme of the invention is as follows: acquiring an original vibration acceleration signal of the whole life cycle of the rolling bearing, inputting the signal into a trained long-short memory network (LSTM) to predict the residual life, inputting the signal into a trained BP neural network to predict the residual error, and adding the results of prediction of the LSTM and the BP neural network to obtain the final residual life.
The present invention will be described in further detail with reference to the accompanying drawings and examples.
Fig. 1 shows a flow chart of the method of the present invention, which specifically includes the following steps:
1. and collecting vibration acceleration signals of a plurality of groups of rolling bearings in the whole life cycle.
2. And processing the original signal in a sliding window mode to obtain a training set of the model, and training the LSTM network model by using the training set.
3. And training a BP neural network for predicting residual errors.
4. And constructing a final residual error correction model, and adding the result of the LSTM network and the result of the BP neural network to obtain a prediction result of the final model.
5. And inputting a signal for predicting the residual life of the rolling bearing to the final model based on residual correction, so as to obtain the residual life at the current moment.
The data acquisition process specifically comprises the following steps:
the method comprises the steps of collecting full-life-cycle vibration data of a plurality of rolling bearings under the same working condition, collecting the data every 10s, setting the sampling frequency to be 25.6kHz, setting the sampling time to be 0.1s, considering that the bearing fails when the measured value of the vibration acceleration reaches a preset threshold value, stopping measuring at the moment, and ending the residual service life of the bearing.
If a certain bearing has 2000 pieces of collected data, the service life is 20000s, when the current sample is the 800 th piece of collected data, the residual life is 12000s, namely 200min, and the label yi during training is the residual life 200.
As shown in fig. 4, which is a schematic diagram of a sliding process, the process of processing an original signal in a sliding window manner to obtain a training set of a model specifically includes:
generally, the input of the LSTM model is 3-dimensional, the acquired vibration acceleration signal is 2-dimensional, so the characteristic input of the LSTM model is obtained in a sliding window mode, and the size of the sliding window is set to be 5-dimensional, namely t, t-1, t-2, t-3 and t-4 are continuously acquired data at 5 moments and used as the characteristic input of the LSTM model at the t moment.
The first 4 pieces of collected data are not enough to slide, but the former data are in a relatively stable state, so that the data can be supplemented by self data, namely the input of the LSTM model at the 1 st moment can be 5 groups of collected data at the 1 st moment.
In this way, the time-of-day characteristics and the corresponding remaining life are combined to form a training set of models.
As shown in fig. 2, the process of training the LSTM network model using the training set specifically includes:
1. building an LSTM network model;
the model is two layers of LSTM networks and one layer of full connection layer.
2. The LSTM optimized parameters are trained using a training set. And obtaining the trained optimal model by adopting a training set 5-fold cross validation mode.
As shown in fig. 3, the process of training the BP neural network for prediction residuals is:
1. firstly, residual errors are obtained, the real residual life corresponding to the vibration acceleration signals of the rolling bearing measured at a certain moment is S, the residual life predicted by the LSTM model is L, and the residual errors Q are S-L.
2. And calculating a residual error as a label of the training data by using the original vibration data of the full life cycle of the rolling bearing as the characteristic of the training data, using the real residual life corresponding to the vibration data of the full life cycle of the rolling bearing and the residual life obtained by prediction of an LSTM model. The features and corresponding labels are combined to form a final training data set.
3. The BP neural network is set to be 1 input layer, 2 hidden layers and 1 output layer.
4. And training the optimal parameters of the BP neural network by using the training set. And obtaining the trained optimal model by adopting a training set 5-fold cross validation mode.
The specific process of constructing the residual error correction model on the basis of the LSTM model comprises the following steps:
1. constructing a final residual error correction model as follows: r — R1(x1) + R2(x2), where R1(x1) is the result of LSTM model prediction and R2(x2) is the result of BP neural network prediction.
2. And solving a final result according to the residual error correction model.
Claims (5)
1. The residual error correction-based residual life prediction method for the rolling bearing is characterized by comprising the following steps of:
1) collecting vibration acceleration signals of the whole life cycle of a rolling bearing;
2) processing the vibration acceleration signal in a sliding window mode to obtain a training set, and training an LSTM network model by using the training set;
3) training a BP neural network model for predicting residual errors;
4) constructing a residual error correction model, and adding the result of the LSTM network model and the result of the BP neural network model to obtain a prediction result of a final model;
5) and inputting a signal for predicting the residual life of the rolling bearing into the residual error correction model to obtain the residual life at the current moment.
2. The residual correction-based rolling bearing residual life prediction method according to claim 1, wherein the input of the LSTM network model is a characteristic of the rolling bearing at time t obtained by a sliding window method; and outputting the service life of the rolling bearing predicted by the LSTM network model at the time t.
3. The residual error correction-based residual life prediction method for the rolling bearing according to claim 1, wherein the input of the BP neural network model is a vibration acceleration signal at the time t, and the output of the BP neural network model is a residual error predicted by the BP neural network model at the time t.
4. The residual error correction-based rolling bearing residual life prediction method according to claim 1 or 3, wherein the residual error algorithm is as follows:
Q＝S-L
and Q is a residual error, S is the real residual life of the rolling bearing corresponding to the vibration acceleration signal of the rolling bearing measured at the time t, and L is the residual life of the rolling bearing predicted by the LSTM model at the time t.
5. The residual correction-based rolling bearing residual life prediction method according to claim 1, wherein the residual correction model is:
R＝R1(x1)+R2(x2)
wherein, R1(x1) is used as L to be the life of the rolling bearing predicted by the LSTM model, R2(x2) is used as Q to be the residual error predicted by the BP neural network, and R is the residual life of the rolling bearing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911290741.7A CN112990524A (en) | 2019-12-16 | 2019-12-16 | Residual error correction-based residual life prediction method for rolling bearing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911290741.7A CN112990524A (en) | 2019-12-16 | 2019-12-16 | Residual error correction-based residual life prediction method for rolling bearing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112990524A true CN112990524A (en) | 2021-06-18 |
Family
ID=76343045
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911290741.7A Pending CN112990524A (en) | 2019-12-16 | 2019-12-16 | Residual error correction-based residual life prediction method for rolling bearing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112990524A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114186500A (en) * | 2022-02-16 | 2022-03-15 | 华中科技大学 | Marine bearing residual life prediction method based on transfer learning and multiple time windows |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0312533A (en) * | 1989-06-10 | 1991-01-21 | Takasago Thermal Eng Co Ltd | Method for estimating remaining life of bearing |
| WO2013160059A1 (en) * | 2012-04-24 | 2013-10-31 | Aktiebolaget Skf | Bearing monitoring method and system |
| CN109543905A (en) * | 2018-11-23 | 2019-03-29 | 西安电子科技大学 | Rolling bearing method for predicting residual useful life based on improved two dimension CNN model |
| CN110097209A (en) * | 2019-03-26 | 2019-08-06 | 朗坤智慧科技股份有限公司 | A kind of equipment deterioration analysis method based on parameter residual error |
| CN110232249A (en) * | 2019-06-17 | 2019-09-13 | 中国人民解放军陆军装甲兵学院 | A kind of rolling bearing method for predicting residual useful life |
-
2019
- 2019-12-16 CN CN201911290741.7A patent/CN112990524A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0312533A (en) * | 1989-06-10 | 1991-01-21 | Takasago Thermal Eng Co Ltd | Method for estimating remaining life of bearing |
| WO2013160059A1 (en) * | 2012-04-24 | 2013-10-31 | Aktiebolaget Skf | Bearing monitoring method and system |
| CN109543905A (en) * | 2018-11-23 | 2019-03-29 | 西安电子科技大学 | Rolling bearing method for predicting residual useful life based on improved two dimension CNN model |
| CN110097209A (en) * | 2019-03-26 | 2019-08-06 | 朗坤智慧科技股份有限公司 | A kind of equipment deterioration analysis method based on parameter residual error |
| CN110232249A (en) * | 2019-06-17 | 2019-09-13 | 中国人民解放军陆军装甲兵学院 | A kind of rolling bearing method for predicting residual useful life |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114186500A (en) * | 2022-02-16 | 2022-03-15 | 华中科技大学 | Marine bearing residual life prediction method based on transfer learning and multiple time windows |
| CN114186500B (en) * | 2022-02-16 | 2022-04-29 | 华中科技大学 | Marine bearing residual life prediction method based on transfer learning and multiple time windows |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109460618B (en) | Rolling bearing residual life online prediction method and system | |
| CN110866314B (en) | Method for predicting residual life of rotating machinery of multilayer bidirectional gate control circulation unit network | |
| CN109814527B (en) | Industrial equipment fault prediction method and device based on LSTM recurrent neural network | |
| CN110232249B (en) | Method for predicting residual life of rolling bearing | |
| JP2020500420A (en) | Yield prediction system and method for machine learning based semiconductor manufacturing | |
| US20240061396A1 (en) | Prediction method for tool remaining life of numerical control machine tool based on hybrid neural model | |
| CN112508053A (en) | Intelligent diagnosis method, device, equipment and medium based on integrated learning framework | |
| CN106127192A (en) | A kind of bearing remaining life Forecasting Methodology based on similarity | |
| CN108961460B (en) | Fault prediction method and device based on sparse ESGP (Enterprise service gateway) and multi-objective optimization | |
| CN114819374A (en) | Regional new energy ultra-short term power prediction method and system | |
| CN115034137A (en) | RVM and degradation model-based two-stage hybrid prediction method for residual life of bearing | |
| CN112990524A (en) | Residual error correction-based residual life prediction method for rolling bearing | |
| CN114216558A (en) | Method and system for predicting remaining life of battery of wireless vibration sensor | |
| CN110850837A (en) | System life analysis and fault diagnosis method based on long-time and short-time memory neural network | |
| CN117216647A (en) | Predictive maintenance and fault diagnosis method and system for vehicle body welding equipment | |
| CN116861503A (en) | Method for constructing digital twin model of power transformer based on big data | |
| CN116431997A (en) | Dynamic threshold calculation method, device and equipment in operation and maintenance monitoring field | |
| CN111914490B (en) | Pump station unit state evaluation method based on depth convolution random forest self-coding | |
| CN115184734A (en) | Power grid line fault detection method and system | |
| CN115203905A (en) | Equipment health assessment method integrating expert experience and intelligent algorithm | |
| CN114819260A (en) | Dynamic generation method of hydrologic time series prediction model | |
| CN114330141A (en) | Fan main shaft bearing service life prediction method based on GRU (generalized regression Unit) super-parameter optimization | |
| CN109242569B (en) | Method and system for analyzing and predicting medium-term and long-term market prices of molybdenum concentrate | |
| CN113465924A (en) | Bearing fault diagnosis method and system based on improved BP neural network | |
| CN113159395A (en) | Deep learning-based sewage treatment plant water inflow prediction method and system |
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 |