CN114298288A - Gearbox fault detection method based on deep migration learning - Google Patents
Gearbox fault detection method based on deep migration learning Download PDFInfo
- Publication number
- CN114298288A CN114298288A CN202210078254.XA CN202210078254A CN114298288A CN 114298288 A CN114298288 A CN 114298288A CN 202210078254 A CN202210078254 A CN 202210078254A CN 114298288 A CN114298288 A CN 114298288A
- Authority
- CN
- China
- Prior art keywords
- data
- gearbox
- fault
- image data
- diagnosis
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 33
- 238000013508 migration Methods 0.000 title claims abstract description 19
- 230000005012 migration Effects 0.000 title claims abstract description 19
- 238000003745 diagnosis Methods 0.000 claims abstract description 30
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 16
- 239000002131 composite material Substances 0.000 claims abstract description 13
- 238000004140 cleaning Methods 0.000 claims abstract description 7
- 150000001875 compounds Chemical class 0.000 claims abstract description 5
- 230000009467 reduction Effects 0.000 claims abstract description 5
- 238000012549 training Methods 0.000 claims abstract description 4
- 238000012545 processing Methods 0.000 claims description 21
- 238000004364 calculation method Methods 0.000 claims description 18
- 238000000034 method Methods 0.000 claims description 11
- 230000002159 abnormal effect Effects 0.000 claims description 4
- 208000004188 Tooth Wear Diseases 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 238000011160 research Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000012546 transfer Methods 0.000 claims description 3
- 238000000354 decomposition reaction Methods 0.000 claims description 2
- 239000000284 extract Substances 0.000 claims 1
- 239000000314 lubricant Substances 0.000 claims 1
- 238000004458 analytical method Methods 0.000 abstract description 3
- 238000012795 verification Methods 0.000 abstract 2
- 239000003921 oil Substances 0.000 description 18
- 238000013135 deep learning Methods 0.000 description 3
- 238000012217 deletion Methods 0.000 description 2
- 230000037430 deletion Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000001816 cooling Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000010687 lubricating oil Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000013526 transfer learning Methods 0.000 description 1
Classifications
-
- 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
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention provides a gearbox fault detection method based on deep migration learning, which comprises the steps of collecting original signal data of a gearbox, carrying out noise reduction and data cleaning on the original signal, extracting by using a feature extractor, obtaining a new image, constructing a model, carrying out result analysis and verification on the model by using training data and verification data, overcoming the technical problem that the type of gearbox fault diagnosis is single, innovatively carrying out cross combination on image data of vibration signals, temperature signals and oil signals of the gearbox to obtain multiple groups of composite image data, finally constructing a composite fault model, migrating the obtained convolutional neural network model to gearbox diagnosis detection data, calculating the migration diagnosis fault rate, wherein the diagnosis rate is between 94 and 100 percent, the high-precision diagnosis is realized, and the fault of the gearbox is quickly detected under the condition of compound fault.
Description
Technical Field
The invention relates to the technical field of gearbox detection, in particular to the technical field of gearbox fault detection based on transfer learning.
Background
The gear box is an important functional part for transmitting motion and allocating speed in the modern equipment manufacturing industry, a gear box system generally comprises four parts, namely a gear, a bearing, a shaft and a box body, faults of the gear box can be divided into mechanical faults, electrical faults and auxiliary system faults, the mechanical faults mainly comprise the gear faults, the bearing faults and the box body faults, and the electrical faults and the auxiliary system faults mainly comprise cooling faults, oil supply faults and sensor faults. Among the three types of faults, although the frequency of occurrence of an electrical fault and an auxiliary system fault is high, the occurrence of the faults is relatively not serious, the handling is convenient, the frequency of mechanical faults is small, the cause is complex, the faults are difficult to find in the early stage, the handling is difficult, and the cost is high. According to statistics, 73% of fault processing time is in processing of mechanical faults, and gears are the parts with multiple mechanical faults.
The gear box comprises a gear box body, a gear box body and a vibration detection method, wherein the gear box body is arranged in the gear box body, the vibration detection method comprises the steps of generating abnormal vibration when a component in the gear box body breaks down, correspondingly changing the amplitude and frequency components of a vibration signal of the gear box body, and containing a large amount of running state information of components inside the gear box body in the gear box body, so that the running state of the gear box body can be effectively reflected by the vibration signal generated by the gear box body, and accurate judgment can be made on the fault without stopping through analysis of the vibration signal of the gear box body. Therefore, vibration detection remains the most widely and effectively used method for diagnosing gearbox faults.
Because the gear box vibration signal data is huge, and deep learning has strong data representation and analysis capability, the fault diagnosis method of deep learning is adopted, in the method for diagnosing the mechanical fault, professional sensor signal data is collected, a fault model and type classification are established by utilizing a deep learning algorithm, compared with the traditional manual detection and detection equipment detection, the fault diagnosis time and the labor and material cost are reduced, and the method has great advantages in the application prospect of the fault diagnosis of the gear box of the wind turbine generator.
However, the type of the fault diagnosis of the gearbox is single at present, the working environment and the fault condition of the gearbox are complex in practice, and the fault condition of the gearbox is not a single component fault and is usually a compound fault, so that various signals in the gearbox need to be considered and fully utilized on the basis of the existing vibration detection, and the gearbox fault detection method based on the deep migration learning is provided.
Disclosure of Invention
Aiming at the technical current situation that a gearbox fault detection method in the prior art is single, the invention aims to provide a gearbox fault detection method based on deep migration learning.
The invention adopts the main technical scheme that:
the invention provides a gearbox fault detection method based on deep migration learning, which comprises the following specific steps of:
(1) collecting original signal data of a gearbox, wherein the original signal data comprises a vibration signal, a temperature signal and an oil signal;
(2) the noise reduction adopts SVD matrix decomposition noise reduction to reduce noise of original vibration signals, remove noise signals, reserve effective vibration signals, convert all effective vibration signals into vibration image data for standby;
(3) performing data cleaning on the collected temperature signal and the collected oil signal, wherein the data cleaning comprises at least one of abnormal value processing, duplicate value deletion and null value processing to obtain temperature image data and oil image data, and the temperature signal comprises at least one of a staged temperature rise value, a same-side temperature difference, a staged uninterrupted temperature rise value and an axle temperature and external temperature difference;
(4) the processed vibration image data, the processed temperature image data and the processed oil image data are combined in a cross mode to obtain multiple groups of composite image data, and a feature extractor is used for extracting to obtain new images;
(5) inputting the obtained new image data into a DarkNet-53 network, performing convolution and pooling calculation on an input data set, extracting three feature maps of different layers from the processed image data, performing convolutional layer processing on the feature maps obtained by sampling at a high layer, performing convolution calculation on two different calculation branches of 3X3 and 1X1 after the convolution calculation is completed, directly outputting a prediction result, splicing the two groups of feature maps through channels, performing convolution processing on the spliced feature maps, repeating the branch calculation processing, and finally outputting the prediction result through a training model value obtained by performing convolution processing on a group of 3X3 and 1X1 to construct a compound fault model.
(6) And comparing the model value obtained by convolution with the theoretical value until the model value is consistent with the theoretical value, obtaining the final convolutional neural network model data, and establishing a convolutional neural network model.
(7) And transferring the obtained convolutional neural network model to the diagnosis and detection data of the gearbox, and calculating the transfer diagnosis fault rate, wherein the diagnosis rate is between 94 and 100 percent, namely, high-precision diagnosis is realized.
Preferably, the feature extractor performs the extracting step by taking the composite image data as a research object, opening a moving window with the same size as the template from the top left corner of the image, multiplying the window image by the template pixel, adding, replacing the pixel brightness value at the center of the window with the calculation result, moving the moving window to the right by one column, performing the same operation, and so on, from left to right, from top to bottom, obtaining a new image, and repeating the steps.
Preferably, the original signal data of the gearbox further comprises wear fault data, pitting fault data, broken tooth fault data, pitting wear mixed fault data, broken tooth wear mixed fault data and normal operation data.
Preferably, the oil signal includes a gearbox oil temperature signal, an oil level signal, and an oil flow signal.
The technical scheme provided by the invention can achieve the following beneficial effects:
the invention provides a gearbox fault detection method based on deep migration learning, which overcomes the technical problem that the type of gearbox fault diagnosis is single, innovatively and crossly combines image data of vibration signals, temperature signals and oil signals of a gearbox to obtain multiple groups of composite image data, finally constructs a composite fault model, compares a model value obtained by convolution processing with a theoretical value until the model value is matched with the theoretical value to obtain final convolutional neural network model data, establishes a convolutional neural network model, migrates the obtained convolutional neural network model into gearbox diagnosis detection data, calculates the migration diagnosis fault rate, and achieves the high-precision diagnosis and the rapid detection of the gearbox fault under the condition of composite fault.
Drawings
None.
Detailed Description
The present invention will be described below by way of examples, but the present invention is not limited to the following examples.
A gearbox fault detection method based on deep migration learning comprises the steps of firstly, collecting original signal data of a gearbox, wherein the original signal data comprises a vibration signal, a temperature signal and an oil signal; decomposing and denoising the original vibration signals by adopting an SVD matrix, removing noise signals, reserving effective vibration signals, and converting all the effective vibration signals into vibration image data for later use; performing data cleaning on the collected temperature signal and the collected oil signal, wherein the data cleaning comprises at least one of abnormal value processing, duplicate value deletion and null value processing to obtain temperature image data and oil image data, and the temperature signal comprises at least one of a staged temperature rise value, a same-side temperature difference, a staged uninterrupted temperature rise value and an axle temperature and external temperature difference; the processed vibration image data, the processed temperature image data and the processed oil image data are combined in a cross mode to obtain multiple groups of composite image data, and a feature extractor is used for extracting to obtain new images; inputting obtained new image data into a DarkNet-53 network, performing convolution and pooling calculation on an input data set, extracting three feature maps of different layers from the processed image data, performing convolutional layer processing on the feature maps obtained by sampling at a high layer, performing convolution calculation on two different calculation branches of 3X3 and 1X1 after the convolution calculation is completed, directly outputting a prediction result, splicing the two groups of feature maps through channels, performing convolution processing on the spliced feature maps, repeating the branch calculation processing work, and finally outputting the prediction result to construct a compound fault model through a training model value obtained by the convolution processing of a group of 3X3 and 1X 1; comparing the model value obtained by convolution with the theoretical value until the model value is consistent with the theoretical value, obtaining the final convolutional neural network model data, and establishing a convolutional neural network model; and transferring the obtained convolutional neural network model to the diagnosis and detection data of the gearbox, and calculating the transfer diagnosis fault rate, wherein the diagnosis rate is between 94 and 100 percent, namely, high-precision diagnosis is realized.
In this embodiment, the feature extractor performs the extraction by taking the composite image data as a research object, opening a moving window with the same size as the template from the top left corner of the image, multiplying the window image by the template pixel, adding the result, replacing the pixel brightness value at the center of the window with the calculation result, moving the moving window to the right in a row, performing the same operation, and so on, from left to right, from top to bottom, to obtain a new image, and repeating the steps.
The original signal data of the gear box further comprise wear fault data, pitting fault data, broken tooth fault data, pitting wear mixed fault data, broken tooth wear mixed fault data and normal operation data.
Wherein the oil signal comprises a gearbox lubricating oil temperature signal, an oil level signal and an oil flow signal.
In conclusion, the gearbox fault detection method based on deep migration learning provided by the invention overcomes the technical problem that the type of gearbox fault diagnosis is single, innovatively and alternately combines the image data of the vibration signal, the temperature signal and the oil signal of the gearbox to obtain multiple groups of composite image data, finally constructs a composite fault model, compares the model value obtained by convolution with the theoretical value until the model value is matched with the theoretical value to obtain the final convolutional neural network model data, establishes the convolutional neural network model, migrates the obtained convolutional neural network model to the gearbox diagnosis detection data, calculates the migration diagnosis fault rate, and realizes high-precision diagnosis when the diagnosis rate is 94-100%.
As described above, the present invention can be preferably implemented, and the above-mentioned embodiments only describe the preferred embodiments of the present invention, and do not limit the scope of the present invention, and various modifications and improvements made to the technical solution of the present invention by those skilled in the art without departing from the design spirit of the present invention should fall within the protection scope determined by the present invention.
Claims (7)
1. A gearbox fault detection method based on deep migration learning is characterized by comprising the following specific steps:
(1) collecting original signal data of a gearbox, wherein the original signal data comprises a vibration signal, a temperature signal and an oil signal;
(2) denoising the original vibration signals, removing the noise signals, reserving effective vibration signals, and converting all the effective vibration signals into vibration image data for later use;
(3) carrying out data cleaning on the collected temperature signals and oil signals;
(4) the processed vibration image data, the processed temperature image data and the processed oil image data are combined in a cross mode to obtain multiple groups of composite image data, and a feature extractor is used for extracting to obtain new images;
(5) inputting obtained new image data into a DarkNet-53 network, performing convolution and pooling calculation on an input data set, extracting three feature maps of different layers from the processed image data, performing convolutional layer processing on the feature maps obtained by sampling at a high layer, performing convolution calculation on two different calculation branches of 3X3 and 1X1 after the convolution calculation is completed, directly outputting a prediction result, splicing the two groups of feature maps through channels, performing convolution processing on the spliced feature maps, repeating the branch calculation processing work, and finally outputting the prediction result to construct a compound fault model through a training model value obtained by the convolution processing of a group of 3X3 and 1X 1;
(6) comparing the model value obtained by convolution with the theoretical value until the model value is consistent with the theoretical value, obtaining the final convolutional neural network model data, and establishing a convolutional neural network model;
(7) and transferring the obtained convolutional neural network model to the diagnosis and detection data of the gearbox, and calculating the transfer diagnosis fault rate, wherein the diagnosis rate is between 94 and 100 percent, namely, high-precision diagnosis is realized.
2. The gearbox fault detection method based on deep migration learning according to claim 1, wherein the method comprises the following steps: the data cleaning comprises at least one of abnormal value processing, repeated value deleting and null value processing, and temperature image data and oil image data are obtained.
3. The gearbox fault detection method based on deep migration learning according to claim 1, wherein the method comprises the following steps: the temperature signal comprises at least one of a staged temperature rise value, a temperature difference on the same side, a staged uninterrupted temperature rise value and a temperature difference between the shaft temperature and the external temperature.
4. The gearbox fault detection method based on deep migration learning according to claim 1, wherein the method comprises the following steps: the characteristic extractor extracts the image data, the composite image data is taken as a research object, a movable window with the same size as the template is opened from the upper left corner of the image, the window image and the template pixel are multiplied and added, the pixel brightness value at the center of the window is replaced by the calculation result, then the movable window moves to the right in a row, the same operation is carried out, the process is repeated, and a new image can be obtained from left to right and from top to bottom.
5. The gearbox fault detection method based on deep migration learning according to claim 1, wherein the method comprises the following steps: the original signal data of the gear box also comprises wear fault data, pitting fault data, broken tooth fault data, pitting wear mixed fault data, broken tooth wear mixed fault data and normal operation data.
6. The gearbox fault detection method based on deep migration learning according to claim 1, wherein the method comprises the following steps: and the noise reduction adopts SVD matrix decomposition noise reduction to reduce the noise of the original vibration signal.
7. The gearbox fault detection method based on deep migration learning according to claim 1, wherein the method comprises the following steps: the oil signal includes a gearbox lubricant temperature signal, an oil level signal, and an oil flow signal.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210078254.XA CN114298288A (en) | 2022-01-24 | 2022-01-24 | Gearbox fault detection method based on deep migration learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210078254.XA CN114298288A (en) | 2022-01-24 | 2022-01-24 | Gearbox fault detection method based on deep migration learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114298288A true CN114298288A (en) | 2022-04-08 |
Family
ID=80978115
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210078254.XA Pending CN114298288A (en) | 2022-01-24 | 2022-01-24 | Gearbox fault detection method based on deep migration learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114298288A (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109376620A (en) * | 2018-09-30 | 2019-02-22 | 华北电力大学 | A kind of migration diagnostic method of gearbox of wind turbine failure |
CN109934222A (en) * | 2019-03-01 | 2019-06-25 | 长沙理工大学 | A kind of insulator chain self-destruction recognition methods based on transfer learning |
CN110220709A (en) * | 2019-06-06 | 2019-09-10 | 北京科技大学 | Fault Diagnosis of Roller Bearings based on CNN model and transfer learning |
CN113865859A (en) * | 2021-08-25 | 2021-12-31 | 西北工业大学 | Multi-scale multi-source heterogeneous information fusion gearbox state fault diagnosis method |
-
2022
- 2022-01-24 CN CN202210078254.XA patent/CN114298288A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109376620A (en) * | 2018-09-30 | 2019-02-22 | 华北电力大学 | A kind of migration diagnostic method of gearbox of wind turbine failure |
CN109934222A (en) * | 2019-03-01 | 2019-06-25 | 长沙理工大学 | A kind of insulator chain self-destruction recognition methods based on transfer learning |
CN110220709A (en) * | 2019-06-06 | 2019-09-10 | 北京科技大学 | Fault Diagnosis of Roller Bearings based on CNN model and transfer learning |
CN113865859A (en) * | 2021-08-25 | 2021-12-31 | 西北工业大学 | Multi-scale multi-source heterogeneous information fusion gearbox state fault diagnosis method |
Non-Patent Citations (9)
Title |
---|
FENG DING, ET AL.: "Fault diagnosis of rolling bearing based on improved CEEMDAN and distance evaluation technique", JOURNAL OF VIBROENGINEERING, 28 February 2017 (2017-02-28), pages 260 - 275 * |
丁锋;秦峰伟;: "小波降噪及Hilbert变换在电机轴承故障诊断中的应用", 电机与控制学报, no. 06, 15 June 2017 (2017-06-15) * |
冯毅雄;赵彬;郑浩;高一聪;杨晨;谭建荣;: "集成迁移学习的轴件表面缺陷实时检测", 计算机集成制造系统, no. 12, 15 December 2019 (2019-12-15) * |
朱云博;周厚强;张燕军;张永强;: "基于多方法结合应用的齿轮箱故障特征提取研究", 山东理工大学学报(自然科学版), no. 05, 25 September 2013 (2013-09-25) * |
毛冠通;洪流;王景霖;: "基于迁移学习的滚动轴承在线故障诊断", 航空科学技术, no. 01, 25 January 2020 (2020-01-25) * |
王靖岳等: "齿轮传动系统故障诊断技术的研究进展", 机械传动, vol. 40, no. 08, 15 August 2016 (2016-08-15), pages 185 - 192 * |
瞿金秀等: "基于多尺度排列熵和支持向量机的轴承故障诊断", 煤矿机械, vol. 39, no. 09, 30 September 2018 (2018-09-30), pages 143 - 146 * |
莫蓓蓓等: "引入Self -Attention 的电力作业违规穿戴智能检测技术研究", 计算机与现代化, no. 2, 29 February 2020 (2020-02-29), pages 115 - 121 * |
陈生等: "动态空间滤波的探讨与实践", 广西师范学院学报, no. 1, 31 March 2006 (2006-03-31), pages 63 - 67 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Jiao et al. | Deep coupled dense convolutional network with complementary data for intelligent fault diagnosis | |
CN111046945B (en) | Fault type and damage degree diagnosis method based on combined convolutional neural network | |
CN112434602B (en) | Fault diagnosis method based on movable common feature space mining | |
Li et al. | Joint attention feature transfer network for gearbox fault diagnosis with imbalanced data | |
Li et al. | Industrial gearbox fault diagnosis based on multi-scale convolutional neural networks and thermal imaging | |
CN111562105B (en) | Wind turbine generator gearbox fault diagnosis method based on wavelet packet decomposition and convolutional neural network | |
CN112305388B (en) | On-line monitoring and diagnosing method for insulation partial discharge faults of generator stator winding | |
CN111198098A (en) | Wind power generator bearing fault prediction method based on artificial neural network | |
CN113656915A (en) | Bearing residual life prediction method based on deep attention network | |
CN116358871A (en) | Rolling bearing weak signal composite fault diagnosis method based on graph rolling network | |
Wang et al. | Fault diagnosis of industrial robots based on multi-sensor information fusion and 1D convolutional neural network | |
CN116842379A (en) | Mechanical bearing residual service life prediction method based on DRSN-CS and BiGRU+MLP models | |
CN114970628A (en) | Rotating part fault diagnosis method based on generation countermeasure network under condition of unbalanced samples | |
Liu et al. | Rolling bearing fault diagnosis of train running gear based on optimized deep residual network | |
CN114298288A (en) | Gearbox fault detection method based on deep migration learning | |
CN111610021B (en) | Early fault detection method for gearbox | |
CN110542557B (en) | Image integral driven machine tool big data periodic fault feature analysis method | |
CN113887586A (en) | FFCNN-SVM transfer learning fault diagnosis method based on feature fusion under small sample | |
Lee et al. | Time series multi-channel convolutional neural network for bearing remaining useful life estimation | |
CN113763244B (en) | Endoscope image super-resolution reconstruction method for diagnosing intelligent pump cavity fault | |
CN116628544A (en) | Fault diagnosis method based on hydraulic winch state on-line monitoring | |
CN114781450B (en) | Train rolling bearing state identification method based on parameter optimization MOMEDA-MIA-CNN | |
CN116227125A (en) | Planetary gear box management system based on digital twin and modeling method | |
He et al. | Fault identification of rotating machinery based on dynamic feature reconstruction signal graph | |
CN114354185A (en) | Gear box fault diagnosis method based on decoupling network |
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 |