CN116049725B - Rotary machine fault diagnosis method based on improved deep learning classification model - Google Patents
Rotary machine fault diagnosis method based on improved deep learning classification model Download PDFInfo
- Publication number
- CN116049725B CN116049725B CN202310314705.XA CN202310314705A CN116049725B CN 116049725 B CN116049725 B CN 116049725B CN 202310314705 A CN202310314705 A CN 202310314705A CN 116049725 B CN116049725 B CN 116049725B
- Authority
- CN
- China
- Prior art keywords
- fault
- signal
- acceleration
- sound
- time
- 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.)
- Active
Links
- 238000003745 diagnosis Methods 0.000 title claims abstract description 33
- 238000000034 method Methods 0.000 title claims abstract description 19
- 238000013135 deep learning Methods 0.000 title claims abstract description 17
- 238000013145 classification model Methods 0.000 title claims abstract description 12
- 230000001133 acceleration Effects 0.000 claims abstract description 63
- 230000004927 fusion Effects 0.000 claims abstract description 41
- 238000010586 diagram Methods 0.000 claims abstract description 37
- 238000012549 training Methods 0.000 claims abstract description 32
- 230000005236 sound signal Effects 0.000 claims abstract description 17
- 238000002372 labelling Methods 0.000 claims abstract description 4
- 230000004913 activation Effects 0.000 claims description 19
- 230000006870 function Effects 0.000 claims description 19
- 230000007246 mechanism Effects 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 6
- 241000282326 Felis catus Species 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 3
- 230000001131 transforming effect Effects 0.000 claims description 3
- 230000006835 compression Effects 0.000 claims description 2
- 238000007906 compression Methods 0.000 claims description 2
- 238000001514 detection method Methods 0.000 abstract description 2
- 238000012360 testing method Methods 0.000 description 4
- 238000000605 extraction Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000012423 maintenance Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention relates to the technical field of rotary machine fault diagnosis, in particular to a rotary machine fault diagnosis method based on an improved deep learning classification model. Extracting sound signals and acceleration signals of each type of fault data from a database; fusing the sound signals and the acceleration signals extracted from each type of fault data to obtain sound acceleration fusion signals of each type of fault data; converting the sound acceleration fusion signals of each type of fault data into a sound acceleration fusion signal time-frequency diagram and labeling fault classification labels; establishing a network model; training and updating parameters of the network model; and inputting the rotating machine fault diagnosis data into the trained improved network model to obtain a final rotating machine fault classification result. Timeliness is guaranteed, the recognition accuracy of an algorithm is improved, the characteristic information of faults is extracted more comprehensively, the equipment detection efficiency is improved, and the accurate classification of the faults of the rotary machine is realized.
Description
Technical Field
The invention relates to the technical field of rotary machine fault diagnosis, in particular to a rotary machine fault diagnosis method based on an improved deep learning classification model.
Background
At present, rotating machinery is widely applied, because the whole equipment is easy to work abnormally, operate unstably, even stop and the like in a relatively severe environment due to long-time operation, deep learning is an emerging research direction in the fields of intelligent monitoring and fault diagnosis of industrial equipment, the rotating machinery is widely used in aerospace and chemical plants, production and operation efficiency of enterprises can be influenced if the rotating machinery breaks down due to the application specificity of the rotating machinery, downtime and maintenance can not be completed in time, casualties of workers can be caused, and even the environment is irreversibly polluted. The traditional rotary machine fault diagnosis method relies on experience of maintenance personnel, has higher shutdown cost and is difficult to realize accurate positioning and diagnosis.
The deep learning related theory is widely applied after being proposed, the popularization of artificial intelligence is promoted while the development is continued, the deep learning related theory is applied to aspects of automatic driving, license plate recognition and the like, the enterprise digital transformation is promoted, the deep learning and the rotating machinery fault diagnosis are closely related, the deep learning related theory is applied to the rotating machinery diagnosis, and a good diagnosis effect is achieved while the safety is improved. However, the existing model has the problems of low efficiency and low precision in fault diagnosis of the rotating machinery, inaccurate feature extraction caused by noise of collected data, and the like, so that the classification of the faults of the rotating machinery is inaccurate, and improvement on a deep learning model for fault diagnosis is needed.
Disclosure of Invention
The invention aims to solve the technical problems that: the fault diagnosis method for the rotary machine overcomes the defects of the prior art and is based on an improved deep learning classification model.
The invention adopts the technical proposal for solving the technical problems that: the rotary machine fault diagnosis method based on the improved deep learning classification model comprises the following steps:
step one: the database stores normal operation data and various fault data of the rotary machine;
step two: from the slaveExtracting sound signal t of each type of fault data from database ip Acceleration signal t iq ；
Step three: for the sound signal t extracted from each type of fault data ip And acceleration signal t iq Fusing to obtain sound acceleration fusion signals W of each type of fault data if ；
Step four: fusing the acoustic acceleration of each type of fault data to a signal W if Time-frequency diagram W converted into sound acceleration fusion signal jf Labeling fault classification labels;
step five: establishing an EfficientNet network model;
step six: training and updating parameters of an EfficientNet network model;
step seven: and inputting the rotating machinery fault diagnosis data into a trained improved EfficientNet network model to obtain a final rotating machinery fault classification result.
The fault data in the database comprises unbalanced fault data, horizontal misalignment fault data, vertical misalignment fault data, inner bearing fault data and outer bearing fault data.
The sound acceleration fusion signal W in the third step if The calculation formula is as follows:
wherein W is if A sound acceleration fusion signal representing i-th type of fault data; f (t) i ) Time domain signal representing i-th type fault data after fusion of sound signal and acceleration signal, and f (t i ) E L2 (R); phi represents the basic wavelet, a represents the frequency parameter, b represents the position parameter, t i Indicating the time corresponding to the fusion signal, i=1, 2, …, n, dt indicates an infinitely small amount of change in time t, - representation ofIs a random variable.
The saidIn the fourth step, the sound acceleration fuses the time-frequency diagram W of the signal jf The calculation formula is as follows:
wherein t is j Represents the time corresponding to the time-frequency diagram data, j represents the fault class and j=1, 2, …, n, phi represents the basic wavelet, a represents the frequency parameter, b represents the position parameter, dt represents the infinitely small amount of time t variation, - representation ofIs a random variable.
And in the fifth step, an Efficient network model is built and initialized, and a multispectral channel attention mechanism and a hard-swish activation function are adopted to replace an original attention mechanism and an original activation function in the Efficient network model, so that an improved Efficient network model is obtained.
The improved attention mechanism of the multispectral channel in the Efficient network model in the step five has the following expression:
Freq＝cat([Freq 0 ,Freq 1 ,…Freq i …,Freq m-1 ])；
wherein Freq represents a multispectral channel vector, freq ε R C Cat denotes a stack of multi-spectral channel vector compression.
The expression of the hard-swish activation function is as follows:
where ReLU6 (x) =min (max (0, x), 6), x is the variable input value of the activation function, and ReLU6 represents the ReLU6 activation function.
The sixth step comprises the following substeps:
6-1: fusing all sound acceleration fusion signals Wj obtained in the step four into a time-frequency chart Wj jf As training set, initializing EfficientNet network model parameters and presettingLearning rate and maximum number of iterations;
6-2: fusion of all sound acceleration signals in training set into time-frequency diagram W jf As an input of an improved EfficientNet network model, outputting a predicted fault classification result corresponding to each time-frequency diagram in the training set;
6-3: calculating a training precision value P through a prediction fault classification result corresponding to the time-frequency diagram and a time-frequency diagram original fault classification label;
6-4: repeating the steps 6-2 and 6-3, training and updating parameters of the Efficient network model, and stopping training when the training reaches the maximum iteration number or the training precision reaches 0.999, so as to obtain a trained improved Efficient network model; otherwise, 6-2 is returned.
The training precision value P in the step 6-4 is expressed as follows:
in the formula, TP represents the same number of the predicted fault classification results of the time-frequency diagram and the original fault classification labels, and FP represents the different number of the predicted fault classification results of the time-frequency diagram and the original fault classification labels.
The seventh step comprises the following substeps:
7-1: receiving an acoustic signal and an acceleration signal of the failed rotary machine;
7-2: fusing the acquired sound signal and the acceleration signal to obtain a sound acceleration fusion signal;
7-3: transforming the sound acceleration fusion signal in the step 7-2 into a sound acceleration fusion signal time-frequency diagram;
7-4: and 7-3, inputting the time-frequency diagram of the sound acceleration fusion signal in the step 7-3 into a trained improved EfficientNet network model, and outputting a fault classification result for obtaining the fault diagnosis data of the rotary machine.
Compared with the prior art, the invention has the following beneficial effects:
1. the method is improved on the basis of an EfficientNet model, a hard-swish activation function is adopted to replace an original activation function, a multispectral channel attention is adopted to replace an SE attention mechanism, more characteristics of different frequency domains of data are identified, timeliness is guaranteed, and the identification accuracy of an algorithm is improved.
2. The problems of insufficient information extraction of a single frequency spectrum in the attention of a channel and the like are solved, the acceleration sensor and the sound sensor are combined, the insufficient signal characteristic extraction is prevented, the multi-sensor combined continuous wavelet transformation technology is adopted, the characteristic information of faults is extracted more comprehensively, the equipment detection efficiency is improved, and the accurate classification of the faults of the rotary machine is realized.
Drawings
FIG. 1 is a flow chart of the system of the present invention.
Detailed Description
As shown in fig. 1, a fault diagnosis method of a rotary machine based on an improved deep learning classification model, a system applied to the method includes a diagnosis terminal, and the rotary machine may be a centrifugal pump, a compressor, a rotary pump, or the like. The rotating machinery is provided with a sound sensor and an acceleration sensor diagnosis terminal, and corresponding information is acquired through the sound sensor and the acceleration sensor on the rotating machinery. The diagnosis terminal is also in communication connection with the fault simulator.
The rotary machine fault diagnosis method based on the improved deep learning classification model comprises the following steps:
step one: the database stores normal operation data and various fault data of the rotary machine; specifically, the diagnosis terminal collects the multivariate time series data of the sensor on the fault simulator, and stores the multivariate time series data in the database according to the data type. The fault data in the database comprises unbalanced fault data, horizontal misalignment fault data, vertical misalignment fault data, inner bearing fault data and outer bearing fault data.
Step two: extracting sound signal t of each type of fault data from database ip Acceleration signal t iq The method comprises the steps of carrying out a first treatment on the surface of the The method mainly comprises the steps that a sound sensor and an acceleration sensor are used for extracting sound signals and acceleration signals for each type of fault data in a database; data are respectively carried out from radial direction, axial direction and tangential direction during acquisitionCollecting;
step three: for the sound signal t extracted from each type of fault data ip And acceleration signal t iq Fusing to obtain sound acceleration fusion signals W of each type of fault data if The method comprises the steps of carrying out a first treatment on the surface of the The sound acceleration fusion signal W in the third step if The calculation formula is as follows:
wherein W is if A sound acceleration fusion signal representing i-th type of fault data; f (t) i ) Time domain signal representing i-th type fault data after fusion of sound signal and acceleration signal, and f (t i ) E L2 (R); phi represents the basic wavelet, a represents the frequency parameter, b represents the position parameter, t i The time corresponding to the fusion signal is indicated, i=1, 2, …, n. dt represents an infinitely small amount of change in time t, - representation ofIs a random variable. In the field of rotary machines, faults generally comprise six categories.
Step four: fusing the acoustic acceleration of each type of fault data to a signal W if Time-frequency diagram W converted into sound acceleration fusion signal jf Labeling fault classification labels; the multi-signal fusion and the signal conversion into the time-frequency diagram are processed by a continuous wavelet conversion method. In the fourth step, the sound acceleration fuses the time-frequency diagram W of the signal jf The calculation formula is as follows:
wherein t is j The time corresponding to the time-frequency diagram data is represented, j represents the fault class and j=1, 2, …, n. Phi denotes the basic wavelet, a denotes the frequency parameter, b denotes the position parameter, dt denotes the infinitely small amount of time t variation, - representation ofIs a random variable.
Step five: establishing an EfficientNet network model; and in the fifth step, an Efficient network model is built and initialized, and a multispectral channel attention mechanism and a hard-swish activation function are adopted to replace an original attention mechanism and an original activation function in the Efficient network model, so that an improved Efficient network model is obtained. The nonlinear activation function has lower cost when the network model deepens the network, and the original swish activation function mainly has larger advantages when being used in the deep network, so that the model is replaced by the hard-swish activation function, and the improvement of precision and the consideration of net influence caused by delay are positive and substantial.
The improved attention mechanism of the multispectral channel in the Efficient network model in the step five has the following expression:
Freq＝cat([Freq 0 ,Freq 1 ,…Freq i …,Freq m-1 ])；
wherein Freq represents a multispectral channel vector, freq ε R C 。
The expression of the hard-swish activation function is as follows:
wherein ReLU6 (x) =min (max (0, x), 6). x is the variable input value of the activation function.
Step six: training and updating parameters of an EfficientNet network model; the sixth step comprises the following substeps:
6-1: fusing all the sound acceleration fusion signals obtained in the step four into a time-frequency diagram W jf As training set, initializing EfficientNet network model parameters, and presetting learning rate and maximum iteration times;
6-2: fusion of all sound acceleration signals in training set into time-frequency diagram W jf As input to the improved EfficientNet network model, the output is a predictive failure classification corresponding to each time-frequency graph in the training setResults;
6-3: calculating a training precision value P through a prediction fault classification result corresponding to the time-frequency diagram and a time-frequency diagram original fault classification label;
6-4: repeating the steps 6-2 and 6-3, training and updating parameters of the Efficient network model, and stopping training when the training reaches the maximum iteration number or the training precision reaches 0.999, so as to obtain a trained improved Efficient network model; otherwise, 6-2 is returned. The training precision value P in the step 6-4 is expressed as follows:
in the formula, TP represents the same number of the predicted fault classification results of the time-frequency diagram and the original fault classification labels, and FP represents the different number of the predicted fault classification results of the time-frequency diagram and the original fault classification labels.
Step seven: and inputting the rotating machinery fault diagnosis data into a trained improved EfficientNet network model to obtain a final rotating machinery fault classification result.
The seventh step comprises the following substeps:
7-1: receiving an acoustic signal and an acceleration signal of the failed rotary machine;
7-2: fusing the acquired sound signal and the acceleration signal to obtain a sound acceleration fusion signal;
7-3: transforming the sound acceleration fusion signal in the step 7-2 into a sound acceleration fusion signal time-frequency diagram;
7-4: and 7-3, inputting the time-frequency diagram of the sound acceleration fusion signal in the step 7-3 into a trained improved EfficientNet network model, and outputting a fault classification result for obtaining the fault diagnosis data of the rotary machine.
And (3) experimental verification:
the database employed in this example consisted of 1951 multivariate time series of sensor acquisitions on alignment-balance-vibration (ABVT) by the Mechanical Fault Simulator (MFS) of SpectraQuest corporation, usa. Wherein 1951 time series includes six different analog states: normal operation data, unbalance faults, horizontal and vertical misalignment faults, inner bearing and outer bearing faults, and acquisition systems are three industrial IMI sensors, 601a01 type radial, axial and tangential accelerometers.
Each sequence was generated within 5 seconds at a sampling rate of 50kHz, with 250000 samples per class. The training is realized in a Pytorch1.8.1 frame, python3.7 and CUDA11.2 environment based on a Linux environment, the precision of an original EfficientNet network test set input by an independent sound signal is 95.83%, the precision of an original EfficientNet network test set input by an independent acceleration sensor signal is 94.91%, the precision of an original EfficientNet network test set input by a fusion signal is 97.22%, and the loss is less 0.0835. The data are input into the improved EfficientNet, and the time-frequency diagram fused with the sound signal and the acquisition signal of the acceleration sensor is compared, so that the accuracy of the test result is improved from 97.22% to 98.75%.
The diagnosis terminal can also be connected with other terminals or mobile terminals (such as a mobile phone, a tablet and the like) to display diagnosis results.
Claims (4)
1. A rotary machine fault diagnosis method based on an improved deep learning classification model, comprising the steps of:
step one: the database stores normal operation data and various fault data of the rotary machine;
step two: extracting sound signal t of each type of fault data from database ip Acceleration signal t iq ；
Step three: for the sound signal t extracted from each type of fault data ip And acceleration signal t iq Fusing to obtain sound acceleration fusion signals W of each type of fault data if ；
Step four: fusing the acoustic acceleration of each type of fault data to a signal W if Time-frequency diagram W converted into sound acceleration fusion signal jf Labeling fault classification labels;
step five: establishing an EfficientNet network model;
step six: training and updating parameters of an EfficientNet network model;
step seven: inputting the rotating machinery fault diagnosis data into a trained improved EfficientNet network model to obtain a final rotating machinery fault classification result;
the fault data in the database comprises unbalanced fault data, horizontal misalignment fault data, vertical misalignment fault data, inner bearing fault data and outer bearing fault data;
the sound acceleration fusion signal W in the third step if The calculation formula is as follows:
wherein W is if A sound acceleration fusion signal representing i-th type of fault data; f (t) i ) Time domain signal representing i-th type fault data after fusion of sound signal and acceleration signal, and f (t i ) E L2 (R); phi represents the basic wavelet, a represents the frequency parameter, b represents the position parameter, t i Indicating the time corresponding to the fusion signal, i=1, 2, …, n, dt indicates an infinitely small amount of change in time t, - representation ofIs a random variable;
in the fourth step, the sound acceleration fuses the time-frequency diagram W of the signal jf The calculation formula is as follows:
wherein t is j Represents the time corresponding to the time-frequency diagram data, j represents the fault class and j=1, 2, …, n, phi represents the basic wavelet, a represents the frequency parameter, b represents the position parameter, dt represents the infinitely small amount of time t variation, - representation ofIs a random variable;
and in the fifth step, an Efficient network model is built and initialized, and a multispectral channel attention mechanism and a hard-swish activation function are adopted to replace an original attention mechanism and an original activation function in the Efficient network model, so that an improved Efficient network model is obtained.
The improved attention mechanism of the multispectral channel in the Efficient network model in the step five has the following expression:
Freq＝cat([Freq 0 ,Freq 1 ,…Freq i …,Freq m-1 ])；
wherein Freq represents a multispectral channel vector, freq ε R C Cat denotes stacking for multi-spectral channel vector compression;
the expression of the hard-swish activation function is as follows:
where ReLU6 (x) =min (max (0, x), 6), x is the variable input value of the activation function, and ReLU6 represents the ReLU6 activation function.
2. The method for diagnosing a rotary machine fault based on an improved deep learning classification model as claimed in claim 1, wherein said step six includes the substeps of:
6-1: fusing all the sound acceleration fusion signals obtained in the step four into a time-frequency diagram W jf As training set, initializing EfficientNet network model parameters, and presetting learning rate and maximum iteration times;
6-2: fusion of all sound acceleration signals in training set into time-frequency diagram W jf As an input of an improved EfficientNet network model, outputting a predicted fault classification result corresponding to each time-frequency diagram in the training set;
6-3: calculating a training precision value P through a prediction fault classification result corresponding to the time-frequency diagram and a time-frequency diagram original fault classification label;
6-4: repeating the steps 6-2 and 6-3, training and updating parameters of the Efficient network model, and stopping training when the training reaches the maximum iteration number or the training precision reaches 0.999, so as to obtain a trained improved Efficient network model; otherwise, 6-2 is returned.
3. The rotary machine fault diagnosis method based on the improved deep learning classification model according to claim 2, wherein the training precision value P in step 6-4 is expressed as follows:
in the formula, TP represents the same number of the predicted fault classification results of the time-frequency diagram and the original fault classification labels, and FP represents the different number of the predicted fault classification results of the time-frequency diagram and the original fault classification labels.
4. The method for diagnosing a rotary machine fault based on the improved deep learning classification model as claimed in claim 2, wherein the seventh step comprises the sub-steps of:
7-1: receiving an acoustic signal and an acceleration signal of the failed rotary machine;
7-2: fusing the acquired sound signal and the acceleration signal to obtain a sound acceleration fusion signal;
7-3: transforming the sound acceleration fusion signal in the step 7-2 into a sound acceleration fusion signal time-frequency diagram;
7-4: and 7-3, inputting the time-frequency diagram of the sound acceleration fusion signal in the step 7-3 into a trained improved EfficientNet network model, and outputting a fault classification result for obtaining the fault diagnosis data of the rotary machine.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310314705.XA CN116049725B (en) | 2023-03-29 | 2023-03-29 | Rotary machine fault diagnosis method based on improved deep learning classification model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310314705.XA CN116049725B (en) | 2023-03-29 | 2023-03-29 | Rotary machine fault diagnosis method based on improved deep learning classification model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116049725A CN116049725A (en) | 2023-05-02 |
| CN116049725B true CN116049725B (en) | 2023-12-29 |
Family
ID=86116628
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310314705.XA Active CN116049725B (en) | 2023-03-29 | 2023-03-29 | Rotary machine fault diagnosis method based on improved deep learning classification model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116049725B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118408743B (en) * | 2024-07-04 | 2024-10-01 | 山东轴研精密轴承有限公司 | Bearing fault detection system and method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106017879A (en) * | 2016-05-18 | 2016-10-12 | 河北工业大学 | Universal circuit breaker mechanical fault diagnosis method based on feature fusion of vibration and sound signals |
| CN114739667A (en) * | 2022-05-10 | 2022-07-12 | 西安交通大学 | Multi-mode information fusion bearing lubrication state monitoring device and method |
| CN114861540A (en) * | 2022-05-12 | 2022-08-05 | 安徽理工大学 | Mine ventilator working state monitoring system based on WT-CNN and harmonic vibration feature fusion |
| CN115238785A (en) * | 2022-07-19 | 2022-10-25 | 上海交通大学宁波人工智能研究院 | Rotary machine fault diagnosis method and system based on image fusion and integrated network |
| CN115687973A (en) * | 2022-10-24 | 2023-02-03 | 江苏科技大学 | Mechanical equipment fault diagnosis method based on multi-signal fusion |
| CN115758212A (en) * | 2022-11-10 | 2023-03-07 | 四川轻化工大学 | Mechanical equipment fault diagnosis method based on parallel network and transfer learning |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110361176B (en) * | 2019-06-05 | 2021-11-19 | 华南理工大学 | Intelligent fault diagnosis method based on multitask feature sharing neural network |
-
2023
- 2023-03-29 CN CN202310314705.XA patent/CN116049725B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106017879A (en) * | 2016-05-18 | 2016-10-12 | 河北工业大学 | Universal circuit breaker mechanical fault diagnosis method based on feature fusion of vibration and sound signals |
| CN114739667A (en) * | 2022-05-10 | 2022-07-12 | 西安交通大学 | Multi-mode information fusion bearing lubrication state monitoring device and method |
| CN114861540A (en) * | 2022-05-12 | 2022-08-05 | 安徽理工大学 | Mine ventilator working state monitoring system based on WT-CNN and harmonic vibration feature fusion |
| CN115238785A (en) * | 2022-07-19 | 2022-10-25 | 上海交通大学宁波人工智能研究院 | Rotary machine fault diagnosis method and system based on image fusion and integrated network |
| CN115687973A (en) * | 2022-10-24 | 2023-02-03 | 江苏科技大学 | Mechanical equipment fault diagnosis method based on multi-signal fusion |
| CN115758212A (en) * | 2022-11-10 | 2023-03-07 | 四川轻化工大学 | Mechanical equipment fault diagnosis method based on parallel network and transfer learning |
Non-Patent Citations (6)
| Title |
|---|
| Fault diagnosis of rolling bearing under time-varying speed conditions based on EfficientNetv2;Haoyang Qu et.al;《Measurement Science and Technology》;第1-14页 * |
| FcaNet: Frequency Channel Attention Networks;Zequn Qin et.al;《arXiv:2012.11879v4 [cs.CV]》;第1-10页 * |
| Identification of Apple Leaf Diseases by Improved Deep Convolutional Neural Networks With an Attention Mechanism;Peng Wang et.al;《Technical Advances in Plant Science》;第1-12页 * |
| 基于声振信号联合的轴承故障诊断方法研究;顾鑫;《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》;第C029-218页 * |
| 基于小波时频图和卷积神经网络的断路器故障诊断分析;鄢仁武;林穿;高硕勋;罗家满;李天建;夏正邦;;振动与冲击(10);第204-211页 * |
| 基于连续小波变换和卷积神经网络的无刷直流电机故障诊断;王骁贤;张保华;陆思良;;机械与电子(06);第31-34页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116049725A (en) | 2023-05-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111721535B (en) | Bearing fault detection method based on convolution multi-head self-attention mechanism | |
| CN109297689B (en) | Large-scale hydraulic machinery intelligent diagnosis method introducing weight factors | |
| Liang et al. | Semisupervised subdomain adaptation graph convolutional network for fault transfer diagnosis of rotating machinery under time-varying speeds | |
| CN103034170B (en) | Numerical control machine tool machining performance prediction method based on intervals | |
| CN108460207A (en) | A kind of fault early warning method of the generating set based on operation data model | |
| CN114993669B (en) | Multi-sensor information fusion transmission system fault diagnosis method and system | |
| CN113008559B (en) | Bearing fault diagnosis method and system based on sparse self-encoder and Softmax | |
| CN116049725B (en) | Rotary machine fault diagnosis method based on improved deep learning classification model | |
| CN110261116A (en) | A kind of Bearing Fault Detection Method and device | |
| CN104571079A (en) | Wireless long-distance fault diagnosis system based on multiple-sensor information fusion | |
| CN116593157A (en) | Complex working condition gear fault diagnosis method based on matching element learning under small sample | |
| CN114266197A (en) | Method for diagnosing equipment fault of hydropower station | |
| CN112305388B (en) | On-line monitoring and diagnosing method for insulation partial discharge faults of generator stator winding | |
| CN113869339A (en) | Deep learning classification model for fault diagnosis and fault diagnosis method | |
| CN111795819B (en) | Gear box fault diagnosis method integrating vibration and current signal collaborative learning | |
| CN103758742A (en) | Plunger pump failure analysis system based on double-class feature fusion diagnosis | |
| CN116881821A (en) | Turbine unit monitoring data anomaly detection method and system based on graph neural network | |
| CN116000131A (en) | Intelligent operation and maintenance method and system for extruder equipment based on data driving | |
| CN117672255B (en) | Abnormal equipment identification method and system based on artificial intelligence and equipment operation sound | |
| CN102096770A (en) | Network-based performance degradation assessment method of rotary machine | |
| CN112177865A (en) | Method for solving marking noise and insufficient marks in fan fault detection | |
| CN116591911B (en) | Intelligent detection operation and maintenance system and method facing to offshore wind turbine generator set | |
| CN118314509A (en) | Device monitoring method, device, storage medium, electronic device, and program product | |
| CN116467592A (en) | Production equipment fault intelligent monitoring method and system based on deep learning | |
| CN114810513A (en) | Wind power generator bearing vibration fault intelligent monitoring system based on 5G communication |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |