CN113537044B - Aircraft engine fault diagnosis method based on STFT and improved DenseNet - Google Patents
Aircraft engine fault diagnosis method based on STFT and improved DenseNet Download PDFInfo
- Publication number
- CN113537044B CN113537044B CN202110794610.3A CN202110794610A CN113537044B CN 113537044 B CN113537044 B CN 113537044B CN 202110794610 A CN202110794610 A CN 202110794610A CN 113537044 B CN113537044 B CN 113537044B
- Authority
- CN
- China
- Prior art keywords
- densenet
- time
- stft
- fault
- channel
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 58
- 238000003745 diagnosis Methods 0.000 title claims abstract description 26
- 230000008569 process Effects 0.000 claims abstract description 16
- 230000001133 acceleration Effects 0.000 claims abstract description 8
- 238000012549 training Methods 0.000 claims description 18
- 238000012360 testing method Methods 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 9
- 238000004458 analytical method Methods 0.000 claims description 7
- 238000000605 extraction Methods 0.000 claims description 7
- 230000005540 biological transmission Effects 0.000 claims description 6
- 238000010606 normalization Methods 0.000 claims description 6
- 230000004927 fusion Effects 0.000 claims description 5
- 238000013519 translation Methods 0.000 claims description 5
- 238000009432 framing Methods 0.000 claims description 4
- 230000009467 reduction Effects 0.000 claims description 4
- 238000001228 spectrum Methods 0.000 claims description 4
- 238000010183 spectrum analysis Methods 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 8
- 230000008901 benefit Effects 0.000 abstract description 3
- 208000031427 Foetal heart rate deceleration Diseases 0.000 abstract description 2
- 238000012795 verification Methods 0.000 abstract description 2
- 239000003638 chemical reducing agent Substances 0.000 description 3
- 230000007797 corrosion Effects 0.000 description 3
- 238000005260 corrosion Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000001066 destructive effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000002405 diagnostic procedure Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000037433 frameshift Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000004575 stone Substances 0.000 description 1
Images
Classifications
-
- 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
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M15/00—Testing of engines
- G01M15/04—Testing internal-combustion engines
- G01M15/12—Testing internal-combustion engines by monitoring vibrations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
- G06F17/141—Discrete Fourier transforms
- G06F17/142—Fast Fourier transforms, e.g. using a Cooley-Tukey type algorithm
-
- 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
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- 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)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Mathematical Analysis (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Optimization (AREA)
- Computational Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Discrete Mathematics (AREA)
- Signal Processing (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- Chemical & Material Sciences (AREA)
- Combustion & Propulsion (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention discloses an aircraft engine fault diagnosis method based on STFT and improved DenseNet, and relates to the problem that the existing method has a poor diagnosis effect on a variable-speed fault data set of an engine in the field of aircraft engine fault diagnosis. The operation conditions of the aircraft engine in the production environment are complex and changeable, and the aircraft engine often undergoes variable speed processes of variable acceleration and variable deceleration, and although the existing method has good effect in a single speed data set, the effect in the variable speed data set is poor; in order to solve the problem, the invention provides an aeroengine fault diagnosis method based on STFT and improved DenseNet; the method comprises the steps of enabling original signals to generate time-frequency images of various faults through STFT, and then classifying the fault images by using an improved DenseNet model; the fault diagnosis method has the advantages that the fault diagnosis effect on the aeroengine is good through sufficient experimental verification. The invention is applied to fault diagnosis of mechanical equipment such as an aircraft engine and the like.
Description
Technical Field
The invention relates to fault diagnosis of mechanical equipment such as an aircraft engine
Background
The aircraft engine is a highly complex and precise machine, is the heart of a spacecraft, is known as the bright pearl on an industrial crown, and is the most complex technical process, the highest technical difficulty and the most expensive part in the aerospace industry. The working environment of the engine is very harsh, and the engine is very easy to have faults, such as corrosion and strike, when the engine runs in the environment with high temperature, high pressure, high stress and strong corrosion for a long time, so that certain parking faults and even destructive damage are brought to the aircraft engine. Therefore, the fault diagnosis of the aircraft engine can effectively improve the working efficiency and the service life of the aircraft engine and reduce the possibility of accidents. However, the aircraft engine has a complicated structure and a large number of parts, so that the failure types are complicated, a series of failure problems including related air passages, oil passages, vibration and the like exist, specific failures such as fouling, thermal deformation, seal fatigue, corrosion and the like are included in the air passage failures, and the vibration failures include specific failures such as fan stall vibration, blade cracks and the like. The key factors caused by different types of fault types are different, for example, the rotating speed of the blades, the actual pressure value of the blades and the like can be used as the factors causing the vibration fault, and the temperature of the air inlet of the engine, the pressure of the air pressure port and the like can be used as the factors causing the air circuit fault.
The fault of the mechanical equipment usually reflects the vibration of the surface of the mechanical equipment, so that the fault diagnosis of the mechanical equipment for a long time mainly utilizes a vibration acceleration sensor to acquire a vibration signal of the surface of the equipment and then utilizes an intelligent algorithm to diagnose the signal. The existing fault diagnosis research of mechanical equipment is based on a bearing fault data set acquired at a single rotating speed such as western university and MFPT, however, in an actual working environment, the operating conditions (different rotating speeds and loads) of the mechanical equipment are constantly changing, especially, an aircraft engine and an accessory subsystem thereof work in a variable rotating speed environment, and the equipment can experience various complex conditions such as repeated acceleration, deceleration, variable acceleration, variable deceleration and the like. But also the resonance of the test bed and the influence of mechanical noise. The difficulty of the data set is high, the existing methods at home and abroad cannot effectively extract and classify the characteristics of the signals, and the extremely poor effect is shown on the Dongan variable-speed data set.
In order to solve the above problems, an aircraft engine fault diagnosis method based on STFT and improved DenseNet is proposed, as shown in fig. 1: the method comprises the steps of carrying out spectrum analysis on an original signal by using STFT to generate a fault time-frequency image, and then training a DenseNet network by using the fault image to classify the fault image; the improved DenseNet network is SEDenseNet (SENet-DenseNet), SEDenseNet is a method of fusing SENet and DenseNet, and an SE module is embedded on the basis of DenseNet-BC, so that not only is the feature reuse and deep information transmission effectively enhanced, but also the recalibration of feature channels is realized, the importance degree of each feature channel is determined through learning, the useful channel features are enhanced according to the weight of each channel, and the channel features which are not useful for the current task are inhibited, and the feature extraction of a time-frequency image is better realized; through a large number of experimental verifications, the method provided by the invention achieves the highest accuracy of 97.32% on a propulsion speed reducer of 'XXX' model of Dongan engine Co., Ltd of China aviation.
Disclosure of Invention
The invention aims to solve the problem that the conventional intelligent detection algorithm cannot realize effective fault diagnosis on an aircraft engine and an accessory subsystem thereof, and provides an aircraft engine fault diagnosis method based on STFT and improved DenseneNet.
The above purpose is realized mainly by the following technical scheme:
the aircraft engine fault diagnosis method based on STFT and improved DenseNet is characterized in that the STFT is used for time-frequency analysis on an original signal to generate a fault time-frequency image, and then an image classification network is used for classification; the method comprises the following steps:
s1, collecting vibration signals of various faults of the engine under the condition of variable-speed operation;
in the data acquisition process, pass through the magnet with vibration acceleration sensor and adsorb on the surface of engine to be connected with the signal acquisition appearance through the connecting wire, use the vibration signal on engine surface of vibration acquisition client software collection of independently developing, in the acquisition process, the workman makes the process of variable rotational speed through the rotational speed and the load of manual control engine, and the software can real-time recording the vibration signal that this in-process produced simultaneously.
S2, constructing an image classification network based on the fusion of SeNet and DenseNet;
embedding an SE module on the basis of DenseNet-BC by utilizing a method of fusing SENet and DenseNet, wherein the SE module mainly comprises four Block modules and four SENet modules, and the four Block modules respectively comprise 3, 3, 6 and 4 DenseLayers; a 1x1 Translation layer structure is arranged behind each Block module for feature dimension reduction, so that not only is feature reuse and deep information transmission effectively enhanced, but also the re-calibration of feature channels is realized, the importance degree of each feature channel is determined through model autonomous learning, useful channel features are enhanced according to the weight of each channel, the channel features with little use on the current task are inhibited, and the feature extraction of a time-frequency image is better realized; finally, classification is realized by using softmax function
S3, performing framing and windowing on the original signal, performing STFT and normalization, and generating a single-channel gray image;
short-time Fourier transform (STFT) is a mathematical transform related to the Fourier transform to determine the frequency and phase of the local area sinusoid of a time-varying signal; the idea is that assuming that an analysis window function g (t) is approximately stable in a short time interval, the window function g (t) is moved on an original time domain vibration signal f (t) so that f (t) g (t) is a stable signal in different finite time widths, and therefore, power spectrums at different moments are calculated; in the short-time Fourier transform process, the length of a window determines the time resolution and the frequency resolution of a spectrogram, the longer the window is, the longer the intercepted signal is, the longer the signal is, the higher the frequency resolution is after Fourier transform, and the worse the time resolution is; conversely, the shorter the window length is, the shorter the intercepted signal is, the worse the frequency resolution is, the better the time resolution is, that is to say, in the short-time fourier transform, the time resolution and the frequency resolution cannot be obtained at the same time, and should be chosen according to specific requirements; after the signals are subjected to frame division and windowing, each frame is subjected to fast Fourier transform and mapped to a pixel representation domain of a single-channel image, and the signals of each frame are superposed to generate a time-frequency image.
S4, making a training set and a testing set;
according to the principle of 7-3 points, the collected 6 fault categories of the east-safety aeroengine under the variable rotating speed include a category of normal data, 70% of samples of each fault are selected as a training set, 30% of samples are selected as a test set, then the training set and the test set are scattered randomly, and the randomness of the samples is increased.
S5, training a model to realize the classification of the fault images;
effects of the invention
The invention mainly aims at the problem that the fault diagnosis accuracy rate of the existing method is low under the variable working condition operating environment of an aircraft engine and an accessory subsystem thereof, and provides an aircraft engine fault diagnosis method based on STFT and improved DenseNet, which is shown in figure 1: the method comprises The steps that through an FFT (fast Fourier transform) powerful time frequency analysis tool, by means of The short-time stability characteristic of The short-time Fourier transform, signals are subjected to framing windowing, fast Fourier transform is carried out and mapped to a pixel representation domain of a single-channel image, a single-channel gray image is generated, finally an image classification network is trained to diagnose a fault image, and The model respectively carries out multiple experiments on American Kesisi storage university Bearing data (CWRU), Italy Metropolis university Rolling Bearing Test bed data set (The Politecnico Di Torino Rolling Bearing Test rig) and China aviation Dongan 'XXX' model reducer 4 channel data sets; the experimental results are shown in fig. 2; on the public data set of single rotating speed of CWRU and university of Metropolis, the method is not away from other parties, especially the accuracy of the CWRU data set, except that MLP is basically over 95%, and the accuracy of the analysis method based on STFT and HHT reaches 100%. On the basis of a variable rotating speed data set, the method based on original signal feature extraction and classification is obviously in a disadvantage, and the accuracy rate of a relatively influential wide convolution shallow model WDCNN on a CWRU data set is only 35%; the accuracy of MSCNN is only 33.24%; the best ALSTM-FCN is expressed in the time sequence classification task, the accuracy rate is slightly higher than the accuracy rate of the ALSTM-FCN, and the accuracy rate of 45% is obtained. Obviously, the model based on time series classification is better than that based on 1DCNN, because the LSTM is added, the time sequence characteristics in the signals can be effectively extracted, and the ALSTM-FCN benefits from the introduction of attention mechanism weighting, and the accuracy is higher than that of the LSTM-FCN by 2 percentage points on average. The method of the invention is obviously better than the diagnostic method based on the original signal, and the average accuracy of 96.4 percent is obtained on a plurality of data sets of Dongan variable rotating speed; also comparing STFT and HHT, it was found that the method based on STFT and the model of the invention was 47 percentage points higher than the average HHT; the effect based on STFT and ResNet34 was even 50 percentage points higher than the average HHT; from the above analysis it follows: the aircraft engine fault diagnosis method based on STFT and improved DenseNet provided by the invention has good effect on the above 6 data sets.
Drawings
Fig. 1 schematic diagram of fault diagnosis based on STFT and modified DenseNet;
FIG. 2 is a summary of the accuracy of the model on each data set;
detailed description of the invention
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without inventive step based on the embodiments of the present invention, are within the scope of protection of the present invention.
Example 1:
the schematic diagram of the method for diagnosing the faults of the aeroengine based on the STFT and the improved DenseNet is shown in figure 1, and the method mainly comprises 2 parts of the STFT and a deep learning classification network:
the model training comprises the steps of:
s1, collecting vibration signals of various faults of the engine under the condition of variable-speed operation;
s2, constructing an image classification network based on the fusion of SeNet and DenseNet;
s3, performing framing and windowing on the original signal, performing STFT and normalization, and generating a single-channel gray image;
s4, making a training set and a testing set;
and S5, training a model to realize the classification of the fault image.
As shown in fig. 1: the method comprises the steps of firstly performing frame windowing on an input signal, performing Fast Fourier Transform (FFT) on each frame, mapping the FFT to a pixel representation domain of a single-channel image, overlapping the images of each frame to generate a fault image, and finally training an SEDenseNet network to classify the fault image.
The following examples illustrate the invention in detail:
1. collecting vibration signals of various faults of an engine under the condition of variable rotating speed operation:
as shown in the upper left part of fig. 1, the experimental vehicle platform of the propulsion retarder of east ampere engine ltd of china aviation, model XXX, is provided with 4 vibration measuring points which are respectively positioned at the left, right, above and below of the equipment; the test bed drives the rotating shaft through the motor to push the speed reducer to rotate; in the data acquisition process, a vibration acceleration sensor is adsorbed on the surface of an engine through a magnet and is connected with a signal acquisition instrument through a connecting line, the vibration acquisition instrument adopts DH5922D of Donghua test, the sensor adopts CA7002A of Huishi stone technology, vibration signals on the surface of the engine are acquired by using self-developed vibration acquisition client software, the sampling frequency is 20khz, in the acquisition process, a worker manually controls the rotating speed and the load of the engine, so that the engine finishes the acceleration and deceleration process within 10S, and simultaneously, the software can record the vibration signals generated in the process in real time; a total of 6 types of fault data including normal state were collected, and 100 samples for each type were collected for 600 samples.
2. Constructing an image classification network based on the fusion of SeNet and DenseNet:
the deep learning model used by the invention is shown in figure 1: embedding an SE module on the basis of DenseNet-BC by utilizing a method of fusing SENet and DenseNet, wherein the SE module mainly comprises four Block modules and four SENet modules, the four Block modules respectively comprise 3, 3, 6 and 4 DenseLayers, and each DenseLayer comprises a 1x1 convolution and a 3x3 convolution; a1 x1 Translation layer structure is arranged behind each Block module for feature dimension reduction, so that not only is feature reuse and deep information transmission effectively enhanced, but also the feature channel is recalibrated, the importance degree of each feature channel is determined through model autonomous learning, useful channel features are enhanced according to the weight of each channel, the channel features which are not useful for the current task are restrained, and the feature extraction of a time-frequency image is better realized.
3. Performing frame windowing on the original signal, performing STFT and normalization, and generating a single-channel gray image:
short-time Fourier transform (STFT) is a mathematical transform related to the Fourier transform to determine the frequency and phase of the local area sinusoid of a time-varying signal; the idea is that assuming that an analysis window function g (t) is approximately stable in a short time interval, the window function g (t) is moved on an original time domain vibration signal f (t) so that f (t) g (t) is a stable signal in different finite time widths, and therefore, power spectrums at different moments are calculated; in the short-time Fourier transform process, the length of a window determines the time resolution and the frequency resolution of a spectrogram, the longer the window is, the longer the intercepted signal is, the longer the signal is, the higher the frequency resolution is after Fourier transform, and the worse the time resolution is; conversely, the shorter the window length is, the shorter the intercepted signal is, the worse the frequency resolution is, and the better the time resolution is, that is, in the short-time fourier transform, the time resolution and the frequency resolution cannot be obtained at the same time, and should be chosen according to specific requirements; as shown in fig. 1: the method selects a Hamming window with the frame length of 25ms and the frame shift of 15ms to window an original signal, respectively performs fast Fourier transform according to each divided frame, and maps frequency spectrum information after the Fourier transform to a pixel representation domain of a single-channel image through normalization processing, wherein the normalization formula is as follows.
And superposing the signals of the frames after the FFT to generate a time-frequency image.
4. Making a training set and a testing set:
the method comprises the steps of collecting 6 fault categories of the Dongan aeroengine under the variable rotating speed, wherein the fault categories comprise a category of normal data, selecting 100 samples of each fault, selecting 70 samples of each fault category as a training set, selecting 30 samples of each fault category as a test set, and then randomly scattering the training set and the test set to increase the randomness of the samples.
5. Training a model to realize classification of fault images:
the input fault image is firstly subjected to convolution of 7x7 and MaxPholing, and then is subjected to four continuous Dense blocks and SENet, wherein the four blocks respectively have 3, 3, 6 and 4 DenseLayers, and each DenseLayer comprises a 1x1 convolution and a 3x3 convolution; each Block module is followed by a Translation layer structure of 1x1, and finally, the classification is realized by global average pooling and the last full connection layer and softmax.
Claims (5)
1. The aircraft engine fault diagnosis method based on STFT and improved DenseNet is characterized in that the method uses STFT to perform spectrum analysis on an original signal to generate a fault time-frequency image, and then uses the fault image to train a DenseNet network to realize classification of the fault image; the improved DenseNet network is SEDenseNet (SEnet-DenseNet), the SEDenseNet is a method of fusing SENet and DenseNet, and an SE module is embedded on the basis of DenseNet-BC, so that not only is the feature reuse and deep information transmission effectively enhanced, but also the recalibration of feature channels is realized, the importance degree of each feature channel is determined through learning, the useful channel features are enhanced according to the weight of each channel, and the channel features which are not useful for the current task are inhibited, and the feature extraction of a time-frequency image is better realized; the method comprises the following steps:
s1, collecting vibration signals of various faults of the engine under the condition of variable-speed operation;
s2, constructing an image classification network based on the fusion of SeNet and DenseNet, embedding an SE module on the basis of DenseNet-BC, wherein the SE module mainly comprises four Block modules and four SEnet modules, and the four Block modules respectively comprise 3, 3, 6 and 4 DenseLayers; a 1x1 Translation layer structure is arranged behind each Block module for feature dimension reduction, so that not only is feature reuse and deep information transmission effectively enhanced, but also the re-calibration of feature channels is realized, the importance degree of each feature channel is determined through model autonomous learning, useful channel features are enhanced according to the weight of each channel, the channel features with little use for the current task are inhibited, and the feature extraction of a time-frequency image is better realized; finally, realizing classification by using a softmax function;
s3, performing framing and windowing on the original signal, performing STFT and normalization, and generating a single-channel gray image;
s4, making a training set and a testing set;
and S5, training a model to realize the classification of the fault image.
2. The STFT and DenseNet-based aircraft engine fault diagnosis method according to claim 1, wherein the principle of collecting vibration signals of the engine under variable-speed operation conditions in step S1 is as follows:
in the data acquisition process, the vibration acceleration sensor is adsorbed on the surface of the engine through a magnet and is connected with the signal acquisition instrument through a connecting wire, vibration signals on the surface of the engine are acquired by using self-developed vibration acquisition client software, in the acquisition process, workers manually control the rotating speed and the load of the engine to enable the engine to complete the process of changing the rotating speed, and meanwhile, the software can record the vibration signals generated in the process in real time.
3. The STFT and DenseNet-based aircraft engine fault diagnosis method according to claim 1, wherein the image classification network based on fusion of SENET and DenseNet in step S2 has the following principles:
embedding an SE module on the basis of DenseNet-BC by using a method of fusing a SEnet and a DenseNet, wherein the SE module mainly comprises four Block modules and four SEnet modules, and the four Block modules respectively comprise 3, 3, 6 and 4 DenseLayers; a 1x1 Translation layer structure is arranged behind each Block module for feature dimension reduction, so that not only is feature reuse and deep information transmission effectively enhanced, but also the re-calibration of feature channels is realized, the importance degree of each feature channel is determined through model autonomous learning, useful channel features are enhanced according to the weight of each channel, the channel features with little use for the current task are inhibited, and the feature extraction of a time-frequency image is better realized; finally, classification is realized by using a softmax function.
4. The STFT and enhanced DenseNet-based aircraft engine fault diagnosis method according to claim 1, wherein the short-time Fourier transform generating time-frequency images in step S3 is based on the following principle:
short-time Fourier transform (STFT) is a mathematical transform related to the Fourier transform to determine the frequency and phase of the local area sinusoid of a time-varying signal; the idea is that assuming that an analysis window function g (t) is approximately stable in a short time interval, the window function g (t) is moved on an original time domain vibration signal f (t) so that f (t) g (t) is a stable signal in different finite time widths, and therefore, power spectrums at different moments are calculated; in the short-time Fourier transform process, the length of a window determines the time resolution and the frequency resolution of a spectrogram, the longer the window is, the longer the intercepted signal is, the longer the signal is, the higher the frequency resolution is after Fourier transform, and the worse the time resolution is; conversely, the shorter the window length is, the shorter the intercepted signal is, the worse the frequency resolution is, and the better the time resolution is, that is, in the short-time fourier transform, the time resolution and the frequency resolution cannot be obtained at the same time, and should be chosen according to specific requirements; after the signals are subjected to frame division and windowing, each frame is subjected to fast Fourier transform and mapped to a pixel representation domain of a single-channel image, and the signals of each frame are superposed to generate a time-frequency image.
5. The STFT and enhanced DenseNet based aircraft engine fault diagnosis method according to claim 1, wherein the method for creating the training set and the test set in step S4 is as follows:
according to the principle of 7-3 points, the collected 6 fault categories of the east-safety aeroengine under the variable rotating speed include a category of normal data, 70% of samples of each fault are selected as a training set, 30% of samples are selected as a test set, then the training set and the test set are scattered randomly, and the randomness of the samples is increased.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110794610.3A CN113537044B (en) | 2021-07-14 | 2021-07-14 | Aircraft engine fault diagnosis method based on STFT and improved DenseNet |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110794610.3A CN113537044B (en) | 2021-07-14 | 2021-07-14 | Aircraft engine fault diagnosis method based on STFT and improved DenseNet |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113537044A CN113537044A (en) | 2021-10-22 |
| CN113537044B true CN113537044B (en) | 2022-08-26 |
Family
ID=78127929
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110794610.3A Expired - Fee Related CN113537044B (en) | 2021-07-14 | 2021-07-14 | Aircraft engine fault diagnosis method based on STFT and improved DenseNet |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113537044B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11859626B2 (en) * | 2022-01-24 | 2024-01-02 | Hamilton Sundstrand Corporation | Incipient compressor surge detection using artificial intelligence |
| TWI812305B (en) * | 2022-06-27 | 2023-08-11 | 國立彰化師範大學 | Fault diagnosis system and fault diagnosis method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108830127A (en) * | 2018-03-22 | 2018-11-16 | 南京航空航天大学 | A kind of rotating machinery fault feature intelligent diagnostic method based on depth convolutional neural networks structure |
| CN110702411A (en) * | 2019-09-23 | 2020-01-17 | 武汉理工大学 | Residual error network rolling bearing fault diagnosis method based on time-frequency analysis |
| CN110991465A (en) * | 2019-11-15 | 2020-04-10 | 泰康保险集团股份有限公司 | Object identification method and device, computing equipment and storage medium |
| CN111178526A (en) * | 2019-12-30 | 2020-05-19 | 广东石油化工学院 | Metamorphic random feature kernel method based on meta-learning |
Family Cites Families (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP5939480B1 (en) * | 2015-12-25 | 2016-06-22 | 富士ゼロックス株式会社 | Terminal device, diagnostic system and program |
| CN106441946B (en) * | 2016-11-15 | 2018-10-19 | 重庆工商大学 | Vehicle hydraulic damper fault recognition method based on vibration signal and system |
| CN107560849B (en) * | 2017-08-04 | 2020-02-18 | 华北电力大学 | Wind turbine generator bearing fault diagnosis method of multichannel deep convolutional neural network |
| CN107421741A (en) * | 2017-08-25 | 2017-12-01 | 南京信息工程大学 | A kind of Fault Diagnosis of Roller Bearings based on convolutional neural networks |
| CN108710892B (en) * | 2018-04-04 | 2020-09-01 | 浙江工业大学 | Cooperative immune defense method for multiple anti-picture attacks |
| CN108875787B (en) * | 2018-05-23 | 2020-07-14 | 北京市商汤科技开发有限公司 | Image recognition method and device, computer equipment and storage medium |
| CN108960257A (en) * | 2018-07-06 | 2018-12-07 | 东北大学 | A kind of diabetic retinopathy grade stage division based on deep learning |
| CN109325516B (en) * | 2018-08-13 | 2021-02-02 | 众安信息技术服务有限公司 | Image classification-oriented ensemble learning method and device |
| CN109190712A (en) * | 2018-09-21 | 2019-01-11 | 福州大学 | A kind of line walking image automatic classification system of taking photo by plane based on deep learning |
| CN109614985B (en) * | 2018-11-06 | 2023-06-20 | 华南理工大学 | Target detection method based on densely connected feature pyramid network |
| CN110046696A (en) * | 2019-04-19 | 2019-07-23 | 电子科技大学 | Residual error neural network method of adjustment based on convolution loop neural network |
| CN110111313B (en) * | 2019-04-22 | 2022-12-30 | 腾讯科技（深圳）有限公司 | Medical image detection method based on deep learning and related equipment |
| CN111413091B (en) * | 2020-04-02 | 2022-05-27 | 天津大学 | Gear box fault diagnosis method under strong noise interference based on data driving |
| CN111914703A (en) * | 2020-07-20 | 2020-11-10 | 哈尔滨工业大学 | Mechanical rotating part fault diagnosis method based on wavelet transformation and transfer learning GoogLeNet |
| CN112378660A (en) * | 2020-10-28 | 2021-02-19 | 西北工业大学 | Intelligent fault diagnosis method for aero-engine bearing based on data driving |
-
2021
- 2021-07-14 CN CN202110794610.3A patent/CN113537044B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108830127A (en) * | 2018-03-22 | 2018-11-16 | 南京航空航天大学 | A kind of rotating machinery fault feature intelligent diagnostic method based on depth convolutional neural networks structure |
| CN110702411A (en) * | 2019-09-23 | 2020-01-17 | 武汉理工大学 | Residual error network rolling bearing fault diagnosis method based on time-frequency analysis |
| CN110991465A (en) * | 2019-11-15 | 2020-04-10 | 泰康保险集团股份有限公司 | Object identification method and device, computing equipment and storage medium |
| CN111178526A (en) * | 2019-12-30 | 2020-05-19 | 广东石油化工学院 | Metamorphic random feature kernel method based on meta-learning |
Non-Patent Citations (1)
| Title |
|---|
| N-DenseNet的城市声音事件分类模型;曹毅等;《西安电子科技大学学报》;20191231(第06期);第15-22+100页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113537044A (en) | 2021-10-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113537044B (en) | Aircraft engine fault diagnosis method based on STFT and improved DenseNet | |
| CN104634571B (en) | A kind of Fault Diagnosis of Roller Bearings based on LCD MF | |
| CN106092565A (en) | A kind of Vibration Analysis method and system thereof | |
| CN109211546B (en) | Rotary machine fault diagnosis method based on noise reduction automatic encoder and increment learning | |
| CN102262215B (en) | Method for detecting stator and rotor air gap eccentric faults of large generator | |
| CN111160167A (en) | Spindle fault classification and identification method based on S-transform deep convolutional neural network | |
| Su et al. | Gear fault identification and classification of singular value decomposition based on Hilbert-Huang transform | |
| CN104215456B (en) | Plane clustering and frequency-domain compressed sensing reconstruction based mechanical fault diagnosis method | |
| CN110044610A (en) | Gear failure diagnosing method | |
| CN114004256A (en) | Fault diagnosis method for manufacturing equipment main bearing based on digital twin body | |
| CN114659790B (en) | Identification method for bearing faults of variable-rotation-speed wind power high-speed shaft | |
| CN110487547A (en) | Fault Diagnosis of Roller Bearings under variable working condition based on vibrorecord and transfer learning | |
| CN112364706A (en) | Small sample bearing fault diagnosis method based on class imbalance | |
| CN111855202A (en) | Gear box fault diagnosis method and system | |
| CN116257747B (en) | Early abnormality detection method for rotor system of fusion kurtosis and convolution self-encoder | |
| CN108956142A (en) | A kind of bearing fault recognition methods | |
| CN113537152A (en) | Flow field state fault detection method based on deep neural network | |
| CN113283028A (en) | Fault diagnosis method for gear of gear box | |
| CN116738314A (en) | Air compressor fault detection method based on neural network | |
| CN112115802A (en) | Crane slewing mechanism gear fault diagnosis method, system and storage medium | |
| Jin et al. | An image recognition method for gear fault diagnosis in the manufacturing line of short filament fibres | |
| Chen et al. | An adversarial learning framework for zero-shot fault recognition of mechanical systems | |
| CN114964769A (en) | Wind power gear box vibration signal fault diagnosis method | |
| CN117109923A (en) | Rolling bearing fault diagnosis method and system | |
| Zhou et al. | Signal de-noising in gear pitting fault identification by an improved singular value decomposition method |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20220826 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |