CN117557865A - PMSM fault diagnosis method and system based on image color characteristics - Google Patents
PMSM fault diagnosis method and system based on image color characteristics Download PDFInfo
- Publication number
- CN117557865A CN117557865A CN202311664830.XA CN202311664830A CN117557865A CN 117557865 A CN117557865 A CN 117557865A CN 202311664830 A CN202311664830 A CN 202311664830A CN 117557865 A CN117557865 A CN 117557865A
- Authority
- CN
- China
- Prior art keywords
- image
- fault diagnosis
- channels
- rgb
- original vibration
- 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
- 238000000034 method Methods 0.000 title claims abstract description 73
- 238000003745 diagnosis Methods 0.000 title claims abstract description 71
- 238000010586 diagram Methods 0.000 claims abstract description 30
- 230000006870 function Effects 0.000 claims abstract description 26
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 20
- 238000009826 distribution Methods 0.000 claims abstract description 18
- 230000004927 fusion Effects 0.000 claims abstract description 10
- 230000001133 acceleration Effects 0.000 claims description 16
- 238000005070 sampling Methods 0.000 claims description 15
- 238000004590 computer program Methods 0.000 claims description 12
- 238000012545 processing Methods 0.000 claims description 9
- 230000008859 change Effects 0.000 claims description 8
- 238000006243 chemical reaction Methods 0.000 claims description 8
- 238000003860 storage Methods 0.000 claims description 6
- 230000001052 transient effect Effects 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 5
- 238000004088 simulation Methods 0.000 claims description 4
- 238000012549 training Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 230000003321 amplification Effects 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000003199 nucleic acid amplification method Methods 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 230000002708 enhancing effect Effects 0.000 claims description 2
- 230000005284 excitation Effects 0.000 claims description 2
- 238000010200 validation analysis Methods 0.000 claims description 2
- 238000000605 extraction Methods 0.000 abstract description 12
- 230000001360 synchronised effect Effects 0.000 abstract description 3
- 230000000694 effects Effects 0.000 description 6
- 238000002474 experimental method Methods 0.000 description 6
- 238000013145 classification model Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 230000007246 mechanism Effects 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 3
- 230000036541 health Effects 0.000 description 3
- 238000012423 maintenance Methods 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 238000004804 winding Methods 0.000 description 2
- 241000533950 Leucojum Species 0.000 description 1
- 108010076504 Protein Sorting Signals Proteins 0.000 description 1
- 238000005299 abrasion Methods 0.000 description 1
- 238000011276 addition treatment Methods 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000005347 demagnetization Effects 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 239000012634 fragment Substances 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000009413 insulation Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/243—Classification techniques relating to the number of classes
- G06F18/2431—Multiple classes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- 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
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Medical Informatics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Multimedia (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a PMSM fault diagnosis method and a PMSM fault diagnosis system based on image color characteristics, wherein the method comprises the steps of obtaining original vibration signals of a motor to be tested under different states and various working conditions; carrying out empirical mode decomposition on the original vibration signal, and solving the eigenmode functions of different levels to obtain signal components of different levels; converting the original vibration signals into symmetrical lattice diagrams of three RGB channels according to signal components of different levels; improving shape distribution and color characteristics of a DenseNet learning image through a fusion scSE module, and constructing a classification network model; and inputting the symmetrical lattice diagram of the RGB three channels into a classification network model, and outputting a fault diagnosis result. The invention has the characteristics of high diagnosis accuracy, non-invasiveness and high robustness, can realize the rapid diagnosis of the turn-to-turn short circuit fault of the permanent magnet synchronous motor, has good generalization capability and anti-interference capability, and ensures faster fault diagnosis and more comprehensive and more accurate fault feature extraction.
Description
Technical Field
The invention belongs to the technical field of motors, and relates to a PMSM fault diagnosis method and system based on image color characteristics.
Background
Permanent Magnet Synchronous Motors (PMSM) are widely applied to new energy automobiles due to the performance advantages of light weight, reliable operation, low noise, high efficiency and the like. Due to the manufacturing defects of the PMSM for the vehicle and the influence of abrasion, deformation, corrosion and other phenomena in the running process of the vehicle, the performance of the PMSM for the vehicle can gradually decline along with the deterioration of the performance of parts, potential safety hazards can be triggered, and even shutdown accidents can occur in serious cases, so that great economic loss is caused. Periodic intermittent operation and frequent starting are easy to cause overheat of the motor winding, and insulation failure is caused. Common fault types of the PMSM for the vehicle include winding short circuit, demagnetizing fault, mechanical fault and the like, wherein the fault characteristics of turn-to-turn short circuit fault, uneven demagnetizing fault and eccentric fault are similar, and the three faults are difficult to distinguish in the prior art. Because acceleration signals of turn-to-turn short circuit faults, uneven demagnetizing faults and eccentric faults have certain similarity and overlapping property, the time domain signals of a single sensor are difficult to distinguish. Therefore, the accurate diagnosis of faults with small signal differences of a single sensor has a certain research significance.
In recent years, a symmetric lattice diagram (SDP) has been widely used in the field of fault diagnosis as an intuitive signal representation method. Unlike the conventional method, SDP can simply convert the original signal into a mirror-symmetrical snowflake image, and the calculation amount is small. The SDP method realizes the identification and discrimination of signals by analyzing the amplitude and frequency difference of images. Thus, the SDP method is considered to be an effective and reliable fault diagnosis method. However, in these methods using SDP, the cases under study are different fault types with large time domain signal differences, and under the condition of small time domain signal differences, the converted SDP has high similarity, which is easy to cause misdiagnosis. The information of the single time domain or time frequency domain dimension cannot completely embody the specific information of the fault, and the information of the two dimensions may need to be combined to make up the limitation of the information of the single dimension and improve the distinguishing degree of the images of different types of motors.
The deep learning method has remarkable progress in fault diagnosis and has the advantage of being capable of automatically learning distinguishing characteristics without relying on expert knowledge. However, most of the deep learning methods at present can only be operated under specific working conditions, and cannot meet the actual requirements. In order to solve this problem, a number of intelligent fault diagnosis methods under non-stationary conditions have appeared in recent years. Although these methods have certain advantages, they still have the following drawbacks due to the stringent requirements of the operating environment that are imposed by conventional time-frequency methods: (1) These algorithms are only applicable in cases where the speed change is small. (2) When the speed is in a transient state, the algorithm performance can be greatly reduced.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a PMSM fault diagnosis method and a PMSM fault diagnosis system based on image color characteristics, which have the characteristics of high diagnosis accuracy, non-invasiveness and high robustness, can realize rapid diagnosis of turn-to-turn short-circuit faults of a permanent magnet synchronous motor, and have good generalization capability and anti-interference capability.
In order to achieve the above purpose, the present invention has the following technical scheme:
in a first aspect, a PMSM fault diagnosis method based on image color features is provided, including:
acquiring original vibration signals of a motor to be tested under different states and various working conditions;
carrying out empirical mode decomposition on the original vibration signal, and solving the eigenmode functions of different levels to obtain signal components of different levels;
converting the original vibration signals into symmetrical lattice diagrams of three RGB channels according to signal components of different levels;
improving shape distribution and color characteristics of a DenseNet learning image through a fusion scSE module, and constructing a classification network model;
and inputting the symmetrical lattice diagram of the RGB three channels into a classification network model, and outputting a fault diagnosis result.
As a preferred solution, the obtaining the original vibration signals of the motor to be tested in different states and under multiple working conditions includes:
constructing a fault simulation experiment table, and presetting inter-turn short circuit faults, uneven demagnetizing faults, eccentric faults and normal motors;
collecting acceleration signal segments of each motor under the following working conditions:
steady state 9 conditions: the rotating speeds are 1000r/min, 1500r/min and 2000r/min respectively, and the loads are no-load, half-load and rated load respectively; transient 6 conditions: the load is acceleration and deceleration working conditions in no-load, half-load and rated load respectively;
each acceleration signal segment is divided into a data subset, which is divided into a training set, a validation set and a test set.
As a preferred solution, the performing empirical mode decomposition on the original vibration signal, and solving the eigenmode functions of different levels to obtain signal components of different levels includes:
carrying out empirical mode decomposition on the preprocessed original vibration signals, and solving eigenvalue functions of different layers;
and indexing all the decomposed signals of the previous three layers of empirical modes after decomposition, and solving the maximum value z of the absolute values of all the sampling points.
As a preferred solution, the converting the original vibration signal according to the signal components of different levels to generate a symmetric lattice diagram of three channels of RGB includes:
representing the time domain features of the original vibration signals by the shape and distribution of the image, and representing the time-frequency domain features by the RGB color features of the image; the signal components of different levels are converted into RGB color features to be displayed on an image, the sizes of the signal components of the first three layers are selected to correspond to the RGB three channels respectively, and the fault features specific to the signals are represented from different time scales.
As a preferred solution, the converting the original vibration signal according to the signal components of different levels to generate a symmetric lattice diagram of three channels of RGB includes:
for time series signals x, x i Is the i-th sampling point value of the time sequence signal x, x i+t Is the adjacent time omega sq Sampling point values after interval t; when the time domain point x i The transformation to polar coordinate space S (r (i), θ (i), φ (i)), is expressed as:
wherein x is min And x max Respectively representing the minimum value and the maximum value of the time domain signal, and the angle theta of the nth mirror symmetry plane n Forming a sector lobe image as a symmetry axis of the polar coordinates; θ (i) and Φ (i) represent the clockwise and counterclockwise rotation angles, respectively, in polar coordinates along the nth plane of mirror symmetry; along with the change of the amplification factor zeta and the time lag factor t, each sampling point of the time domain signal is intuitively displayed in the polar coordinate image to present different images;
solving the RGB three-channel characteristic sequence T by utilizing the maximum value z of the absolute values of all sampling points R (i)T G (i)T B (i) The method comprises the following steps:
wherein, IMF1 i For the first layer empirical mode decomposition component, corresponding to the ith value, IMF1 i+t Decomposing components for the first layer of empirical mode, wherein the i+t number is corresponding to the first layer of empirical mode;
scaling the RGB three channel signature sequence to within the range of 0-255 according to the following formula:
wherein Round () is a rounding function; k=R, G, B represent image color channels, R, G and B channels correspond to the front three-stage intrinsic mode components respectively; t (T) k (i) Representing an RGB three-channel feature sequence;
and displaying the color characteristics corresponding to all points on the symmetrical lattice diagram.
As a preferable scheme, the step of improving the shape distribution and color characteristics of the DenseNet learning image through the fusion of the scSE module and constructing a classification network model is carried out, the extraction of the importance degree on the channel and the space is carried out on the input characteristic image respectively, and after the addition treatment is carried out, the characteristic subgraphs with important channels and important space characteristics are enhanced and excited at the same time, so that the network learning of more meaningful characteristic information is promoted.
As a preferable scheme, the method for improving the shape distribution and the color characteristics of the DenseNet learning image by fusing the scSE module comprises the following steps:
input feature map p= [ P ] 1 ,p 2 ,…,p N ]Wherein the channelL and W are the height and width of the feature map, respectively;
the global spatial features are embedded into the vector v by a global pooling layer, wherein,c is the number of channels;
the obtained vector v is respectively omega by weight 1 And omega 2 Is subjected to a ReLU activation function and Sigmoid normalization processing in sequence to obtain an ith channel p i Degree of feature importance of (a)Wherein (1)>The value of (2) is
The number of the through channels is C, and the weight is v sq The input feature map P is channel compressed by a 1×1 convolution block, the number of output channels is 1, and the feature map fq=ω with a size of lxw sq *P；
Normalizing the obtained feature map fq through Sigmod to obtain the spatial information importance degree sigma (fq (i, j)) of each spatial position (i, j) in the feature map, and enhancing the important spatial position features;
for a pair ofAnd sigma (fq (i, j)) to obtain a feature sub-graph P'.
In a second aspect, a PMSM fault diagnosis system based on image color features is provided, including:
the original vibration signal acquisition module is used for acquiring original vibration signals of the motor to be tested under different states and various working conditions;
the empirical mode decomposition module is used for performing empirical mode decomposition on the original vibration signal and solving the intrinsic mode functions of different levels to obtain signal components of different levels;
the symmetrical lattice diagram conversion module is used for converting the original vibration signals into symmetrical lattice diagrams of three RGB channels according to signal components of different levels;
the classifying network model building module is used for improving the shape distribution and color characteristics of the DenseNet learning image through the fusion scSE module to build a classifying network model;
and the output module is used for inputting the symmetrical lattice diagrams of the RGB three channels into the classification network model and outputting fault diagnosis results.
In a third aspect, there is provided an electronic device comprising: a memory storing at least one instruction; and a processor executing the instructions stored in the memory to implement the image color feature-based PMSM fault diagnosis method.
In a fourth aspect, a computer readable storage medium is provided, where a computer program is stored, which when executed by a processor, implements the PMSM fault diagnosis method based on image color characteristics.
Compared with the prior art, the invention has at least the following beneficial effects:
in order to improve the distinguishing capability of similar fault types, the invention provides a signal-image conversion method based on empirical mode decomposition-symmetric point mode (EMD-SDP), which displays the time-frequency domain characteristics of the EMD on the SDP image in a color form and reduces the similarity between classes of SDP images generated by similar faults. In addition, the invention also provides a classification network model for improving the DenseNet convolutional neural network based on the fusion of the scSE attention mechanism, so that the learning capacity of the network on important channels and spatial characteristics in the feature map is enhanced, and the fault diagnosis effect is optimized. Experimental results of different types of PMSMs under various conditions verify that the performance of the proposed method is superior to other baseline methods. The method provided by the invention has higher diagnosis accuracy under the constant-speed and constant-load working condition, has certain generalization capability under the variable-speed and constant-load working condition, overcomes the limitation of lower fault accuracy of a single sensor with smaller diagnosis signal difference, provides theoretical guidance and actual reference for reducing the maintenance cost of a complex motor system and improving the service quality, and ensures faster fault diagnosis and more comprehensive and more accurate fault feature extraction.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present invention, and that other related drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a PMSM fault diagnosis method based on image color characteristics according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of an EMD-RGB feature enhancement SDP conversion process according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a DenseNet network architecture incorporating scSE modules in accordance with an embodiment of the present invention;
FIG. 4 is an EMD-SDP image of different loads at a speed of 1000r/min according to an embodiment of the present invention;
fig. 5 is an image of EMD-SDP at different speeds at rated load of an embodiment of the present invention;
fig. 6 is an EMD-SDP image of a normal motor during idle acceleration and deceleration in accordance with an embodiment of the present invention;
FIG. 7 is an EMD-SDP image of an inter-turn short circuit motor during half-load acceleration and deceleration in accordance with an embodiment of the present invention;
FIG. 8 is an EMD-SDP image of an eccentric motor during rated load acceleration and deceleration in accordance with an embodiment of the present invention;
FIG. 9 is a confusion matrix plot of steady state datasets obtained by a deep learning method in accordance with an embodiment of the present invention;
FIG. 10 is a histogram of diagnostic accuracy for constant load acceleration and deceleration using various methods in accordance with embodiments of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, one of ordinary skill in the art may also obtain other embodiments without undue burden.
The embodiment of the invention provides a PMSM fault diagnosis method based on image color characteristics, which ensures faster fault diagnosis and more comprehensive and more accurate fault characteristic extraction, has better performance under steady-state and variable-speed transient working conditions, has overall performance superior to other methods, and can be effectively used for fault diagnosis during speed change.
Firstly, it should be noted that, in the field of fault diagnosis, in order to obtain as much information as possible, a higher sampling rate is generally adopted, so that fault data in the following embodiments of the present invention are all measured at 10 kHz.
Example 1
Referring to fig. 1, a PMSM fault diagnosis method based on image color features includes the following steps:
s1: obtaining vibration signals of motors in different states under various working conditions through experiments, wherein the vibration signals comprise four motor states including a health state, a turn-to-turn short circuit fault, a local demagnetizing fault and an eccentric fault, various rotating speed and load working conditions and vibration acceleration signals of motors in constant-load variable-speed working conditions;
s2: EMD is carried out on the preprocessed vibration signals, different-level eigen mode functions are solved, the special fault characteristics of the signals can be represented from different time scales, and the method is suitable for analyzing nonlinear and non-stationary signal sequences and has high signal-to-noise ratio;
s3: and converting the original vibration signal to generate an RGB-SDP image, wherein the time domain features of the original signal are represented by the shape and distribution of the image, and the time-frequency domain features are represented by the RGB color features of the image. The IMFs of different levels are converted into RGB color features and displayed on an image, the sizes of the first three IMFs are selected to correspond to the three RGB channels respectively, and the fault features specific to signals can be represented from different time scales, as shown in fig. 2.
S4: the shape distribution and color characteristics of a DenseNet learning image are improved through the fusion of the scSE module, a classification network model is constructed, the characteristic extraction capacity of a network is enhanced by the improvement method, the attention of the network to important characteristics of the image is improved through a weighting strategy, the representation capacity of a backbone network to important space and channel characteristics is enhanced, and the method can be better suitable for various complex scenes;
s5: s2 and S3 are executed on vibration signals of the motor to be tested, and an image is input into the constructed classification model by utilizing the step S4 to diagnose motor faults;
in one possible implementation, step S1 specifically includes:
s11: a PMSM fault simulation experiment table is constructed, and inter-turn short circuit faults, uneven demagnetization faults, eccentric faults and normal motors are preset in advance;
s12: collecting acceleration signal fragments under the common working conditions of each motor and the steady-state 9 working conditions (the rotating speeds are respectively 1000r/min, 1500r/min and 2000r/min, the loads are respectively no-load, half-load and rated load) and the transient 6 working conditions (the acceleration and deceleration working conditions when the loads are respectively no-load, half-load and rated load), as shown in figures 4 to 8;
s13: dividing each acceleration signal segment into 400 data subsets, and dividing all the data subsets into three parts of training sets, verification sets and test sets;
in one possible implementation, step S2 specifically includes:
s21: EMD is carried out on the preprocessed vibration signals, and eigen mode functions of different levels are solved;
s22: and indexing all signals of the three layers of EMDs before decomposition, and solving the maximum value z of the absolute values of all sampling points.
In one possible implementation, step S3 specifically includes:
s31: for time series signals x, x i Is the i-th sample point value of the signal x, x i+t Is the adjacent time omega sq Sample point value of t after interval. According to the principle of SDP, when the time domain point x i The transformation to polar coordinate space S (r (i), θ (i), φ (i)), can be expressed as:
wherein x is min And x max Respectively representing the minimum value and the maximum value of the time domain signal, and the angle theta of the nth mirror symmetry plane n As symmetry axes of polar coordinates, a sector lobe image is formed. θ (i) and Φ (i) represent the clockwise and counterclockwise rotation angles, respectively, in polar coordinates along the nth plane of mirror symmetry. With the change of the amplification factor ζ and the time lag factor t, each sampling point of the time domain signal can be intuitively displayed in the polar coordinate image to present different images.
S32: due to the use of x in the conversion process of SDP images i And x i+t Values of the successive sampling points, IMF after EMD conversion i And IMF (inertial measurement unit) i+t Corresponding to this. Three-layer EMD is carried out on all acquired signals, and all decomposed signalsIndexing, solving the maximum value z of absolute values of all sampling points, and solving the RGB three-channel characteristic sequence T R (i)T G (i)T B (i) The method comprises the following steps:
wherein, IMF1 i For the ith value corresponding to the first layer empirical mode decomposition component, IMF1 i+t The i+t-th value corresponds to the first-layer empirical mode decomposition component.
S33: the RGB three channel feature sequence is scaled in a proportion to within the range of 0-255. The specific processing mode is as follows:
wherein Round () is a rounding function; k=r, G, B represent image color channels, R, G and B channels correspond to the first three-stage eigenmode components, respectively; t (T) k (i) Representing an RGB three-channel feature sequence.
S34: and displaying the color features corresponding to all the points on the SDP image.
In a possible implementation manner, the scSE module in step S4 extracts the channel and the spatial importance degree of the input feature map respectively, and performs addition processing, so that the obtained feature subgraphs with high importance (i.e. having important channels and important spatial features at the same time) are subjected to stronger excitation, and network learning is promoted to learn more meaningful feature information. As shown in fig. 3, the specific steps of extracting and optimizing the features by the module are as follows:
s41: input feature map p= [ P ] 1 ,p 2 ,…,p N ](wherein the channelL and W are the height and width of the feature map, respectively
S42: embedding global spatial features into vectors through global pooling layerIn v (whereinC is the number of channels
S43: the obtained vector v is respectively omega by weight 1 And omega 2 And sequentially carrying out a ReLU activation function (delta (·)) and a Sigmoid normalization process (sigma (·)) to obtain an ith channel p i Degree of feature importance of (a)Wherein->The value of +.>
S44: the number of the through channels is C, and the weight is omega sq The input feature map P is channel compressed by a 1×1 convolution block, the number of output channels is 1, and the feature map of size l×w is fq=ω sq *P；
S45: and (3) normalizing the obtained feature map fq by Sigmod (sigma ()) to obtain the importance degree (sigma (fq (i, j))) of the spatial information of each spatial position (i, j) in the feature map so as to enhance the important spatial position feature.
S46: for a pair ofAnd sigma (fq (i, j)) to obtain a feature subgraph P';
in one possible implementation, step S5 specifically includes:
s51: EMD is carried out on the vibration signals to be tested, and intrinsic mode functions of different levels are solved;
s52: converting the vibration signals to be detected into SDP images, converting IMFs of different layers into color features, and displaying the color features on the SDP images;
s53: and inputting an SDP image with improved EMD-RGB characteristics converted by the vibration signal to be detected, inputting a fused scSE attention mechanism improved DenseNet classification network model, and outputting a diagnosis result of the signal to be detected.
According to the invention, vibration signals of motors in different states under various working conditions are obtained through experiments; EMD is carried out on the preprocessed vibration signals, and eigen mode functions of different levels are solved; and converting the original vibration signal to generate an RGB-SDP image, wherein the time domain features of the original signal are represented by the shape and distribution of the image, and the time-frequency domain features are represented by the RGB color features of the image. The IMFs of different levels are converted into RGB color features and displayed on an image, the sizes of the first three IMFs are selected to correspond to the RGB three channels respectively, and the special fault features of the signals can be represented from different time scales. The characteristic features of the fault signals can be more comprehensively represented by combining the time domain features and the time-frequency domain features, so that further feature extraction and model construction are facilitated; the shape distribution and color characteristics of a DenseNet learning image are improved through the fusion of the scSE module, a classification network model is constructed, the characteristic extraction capacity of a network is enhanced by the improvement method, the attention of the network to important characteristics of the image is improved through a weighting strategy, the representation capacity of a backbone network to important space and channel characteristics is enhanced, and the method can be better suitable for various complex scenes; and (3) executing the steps S2 and S3 on the vibration signal of the motor to be tested, and diagnosing the motor faults by utilizing the classification model constructed in the step S4. The PMSM fault diagnosis method based on the image color features ensures faster fault diagnosis and more comprehensive and more accurate fault feature extraction, has better performance under steady-state and variable-speed transient working conditions, has overall performance superior to other methods, and can be effectively used for fault diagnosis during speed change. The method overcomes the limitation of low fault accuracy of small signal difference of single sensor diagnosis, and provides theoretical guidance and practical reference for reducing the maintenance cost of a complex motor system and improving the service quality.
Example two
Based on the same inventive concept as the PMSM fault diagnosis method based on the image color feature in the first embodiment, the present invention further provides a PMSM fault diagnosis system based on the image color feature, which includes:
the original vibration signal acquisition module is used for acquiring original vibration signals of the motor to be tested under different states and various working conditions;
the empirical mode decomposition module is used for performing empirical mode decomposition on the original vibration signal and solving the intrinsic mode functions of different levels to obtain signal components of different levels;
the symmetrical lattice diagram conversion module is used for converting the original vibration signals into symmetrical lattice diagrams of three RGB channels according to signal components of different levels;
the classifying network model building module is used for improving the shape distribution and color characteristics of the DenseNet learning image through the fusion scSE module to build a classifying network model;
and the output module is used for inputting the symmetrical lattice diagrams of the RGB three channels into the classification network model and outputting fault diagnosis results.
Example III
Based on the same inventive concept as the PMSM fault diagnosis method based on the image color feature in the first embodiment, the present invention further provides an electronic device, including: a memory storing at least one instruction; and a processor executing instructions stored in the memory to implement the PMSM fault diagnosis method based on image color characteristics according to the first embodiment, specifically as follows:
the device comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, and is characterized in that the processor executes the program to realize the following steps:
obtaining vibration signals of PMSM with health state, turn-to-turn short circuit fault, local demagnetizing fault and eccentric fault under various rotating speed and load working conditions;
EMD is carried out on the preprocessed vibration signals, and eigen mode functions of different levels are solved;
converting the original vibration signals to generate SDP images, converting IMFs of different layers into color features, and displaying the color features on the SDP images;
improving a DenseNet learning image data set by fusing a scSE attention mechanism to construct a classification network model;
and executing the step S2 and the step S3 on vibration signals of the motor to be tested, inputting pictures into the constructed classification model by utilizing the step S4, and diagnosing motor faults.
Example IV
Based on the same inventive concept as the PMSM fault diagnosis method based on the image color feature in the first embodiment, the present invention further provides a computer readable storage medium storing a computer program, which when executed by a processor, implements the PMSM fault diagnosis method based on the image color feature of the first embodiment:
obtaining vibration signals of motors in different states under various working conditions through experiments, wherein the vibration signals comprise four motor states including a health state, a turn-to-turn short circuit fault, a local demagnetizing fault and an eccentric fault, and various rotating speed and load working conditions;
EMD is carried out on the preprocessed vibration signals, and eigen mode functions of different levels are solved;
converting the original vibration signals to generate SDP images, converting IMFs of different layers into color features, and displaying the color features on the SDP images;
improving a DenseNet learning image data set by fusing a scSE attention mechanism to construct a classification network model;
and executing the step S2 and the step S3 on vibration signals of the motor to be tested, inputting pictures into the constructed classification model by utilizing the step S4, and diagnosing motor faults.
In the embodiment of the invention, the PMSM fault experiment platform is used for collecting vibration data, so that the experiment verification of the related theory is completed. The motor fault experiment table mainly comprises a motor to be detected, a vibration sensor, a positioning platform, a digital signal acquisition device, an encoder, a direct current power supply and a computer, wherein motor parameters are shown in table 1.
Table 1 shows motor parameters
| Rated power | Number of poles | Voltage (V) | Electric current | Frequency of | Rated rotational speed |
| 750w | 4 | 380V | 1.35A | 50Hz | 3000rpm |
The fault motor mainly comprises an inter-turn short circuit fault motor, an uneven demagnetizing fault motor and an eccentric fault motor.
As shown in fig. 9 and fig. 10, in summary, the present invention provides a PMSM fault diagnosis method based on image color features, aiming at the problem that the conventional diagnosis method cannot meet the actual requirements when the working conditions of the PMSM change. The fault diagnosis method has higher precision under the constant-speed and constant-load working condition and better generalization capability under the variable-speed and constant-load working condition, and meanwhile, the diagnosis method provided by the invention is non-invasive and can be popularized and applied to the state monitoring and diagnosis of the industrial motor. The method overcomes the limitation of low fault accuracy of small signal difference of single sensor diagnosis, and provides theoretical guidance and practical reference for reducing the maintenance cost of a complex motor system and improving the service quality.
The invention provides an EMD-RGB feature-based SDP time sequence signal image conversion method, which respectively corresponds the sizes of the first three IMFs to R, G and B channel intensities of a color image. The time domain features of the original signal are represented by the shape and distribution of the image, and the time-frequency domain features are represented by the RGB color features of the image. Compared with a gray level image, the method has the advantages that the complete original time domain signal information is reserved, meanwhile, the signal characteristics of a time domain are reflected, the difference of images among classes is increased, and the fault type with small time domain signal difference can be distinguished. The time domain and the time frequency domain features are combined, so that the characteristic features of the fault signal can be more comprehensively represented, and further feature extraction and model construction are facilitated.
According to the invention, the scSE attention module is embedded into the DenseNet network, so that the characteristic extraction effect of the network is improved. The attention of the network to important features of the image is improved through the weighting strategy, the representation capability of the backbone network to important space and channel features is enhanced, and the problem that the extraction effect of the network to complex high-level features is poor is effectively solved. By calculating the importance of the channel and the spatial feature in the feature map, the learning capability of the network to the important channel and the spatial feature in the feature map is enhanced, the fault diagnosis effect is optimized, and the distribution of the diagnostic model to the image and the extraction capability of the color feature are improved.
The invention constructs a PMSM fault simulation test bed, tests different types of PMSM under constant-speed constant-load working condition and variable-speed constant-load working condition, and verifies the proposed method. According to the method, multiple rotational speeds and load experimental data under the constant-speed constant-load working condition are used for training, and a classification model is constructed. The fault diagnosis result shows that the method has the highest diagnosis accuracy. The method is also suitable for the constant load acceleration and constant load deceleration working conditions of the motor, can be effectively used for fault diagnosis during speed change, and has certain generalization performance.
The diagnosis idea provided by the invention is suitable for other time sequence imaging methods, has good universality and good application effect. The fault diagnosis effect by using a single dimension of the signal is not ideal, and the diagnosis precision can be greatly improved by using the characteristics of various dimensions of the signal. The convolutional neural network reduces the complexity of a network model through 3 strategies of local receptive field, weight sharing and downsampling, and has the characteristic of invariance to forms such as translation, rotation, scaling and the like.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be apparent to those skilled in the art that various modifications and variations can be made to the present invention without departing from the spirit or scope of the invention. Thus, it is intended that the present invention also include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.
Claims (10)
1. A PMSM fault diagnosis method based on image color characteristics, comprising:
acquiring original vibration signals of a motor to be tested under different states and various working conditions;
carrying out empirical mode decomposition on the original vibration signal, and solving the eigenmode functions of different levels to obtain signal components of different levels;
converting the original vibration signals into symmetrical lattice diagrams of three RGB channels according to signal components of different levels;
improving shape distribution and color characteristics of a DenseNet learning image through a fusion scSE module, and constructing a classification network model;
and inputting the symmetrical lattice diagram of the RGB three channels into a classification network model, and outputting a fault diagnosis result.
2. The PMSM fault diagnosis method based on the image color characteristics according to claim 1, wherein the obtaining the original vibration signals of the motor to be tested under different states and under various working conditions comprises:
constructing a fault simulation experiment table, and presetting inter-turn short circuit faults, uneven demagnetizing faults, eccentric faults and normal motors;
collecting acceleration signal segments of each motor under the following working conditions:
steady state 9 conditions: the rotating speeds are 1000r/min, 1500r/min and 2000r/min respectively, and the loads are no-load, half-load and rated load respectively; transient 6 conditions: the load is acceleration and deceleration working conditions in no-load, half-load and rated load respectively;
each acceleration signal segment is divided into a data subset, which is divided into a training set, a validation set and a test set.
3. The PMSM fault diagnosis method based on image color characteristics according to claim 1, wherein the performing empirical mode decomposition on the original vibration signal, and solving different levels of eigenmode functions to obtain different levels of signal components includes:
carrying out empirical mode decomposition on the preprocessed original vibration signals, and solving eigenvalue functions of different layers;
and indexing all the decomposed signals of the previous three layers of empirical modes after decomposition, and solving the maximum value z of the absolute values of all the sampling points.
4. The PMSM fault diagnosis method based on image color characteristics according to claim 1, wherein the converting the original vibration signal according to the different level signal components to generate the symmetric lattice diagram of the RGB three channels comprises:
representing the time domain features of the original vibration signals by the shape and distribution of the image, and representing the time-frequency domain features by the RGB color features of the image; the signal components of different levels are converted into RGB color features to be displayed on an image, the sizes of the signal components of the first three layers are selected to correspond to the RGB three channels respectively, and the fault features specific to the signals are represented from different time scales.
5. The PMSM fault diagnosis method based on image color characteristics according to claim 4, wherein the converting the original vibration signal according to the different level signal components to generate the symmetric lattice diagram of the RGB three channels comprises:
for time series signals x, x i Is the i-th sampling point value of the time sequence signal x, x i+t Is the adjacent time omega sq Sampling point values after interval t; when the time domain point x i The transformation to polar coordinate space S (r (i), θ (i), φ (i)), is expressed as:
wherein x is min And x max Respectively representing the minimum value and the maximum value of the time domain signal, and the angle theta of the nth mirror symmetry plane n Forming a sector lobe image as a symmetry axis of the polar coordinates; θ (i) and Φ (i) represent the clockwise and counterclockwise rotation angles, respectively, in polar coordinates along the nth plane of mirror symmetry; along with the change of the amplification factor zeta and the time lag factor t, each sampling point of the time domain signal is intuitively displayed in the polar coordinate image to present different images;
solving the RGB three-channel characteristic sequence T by utilizing the maximum value z of the absolute values of all sampling points R (i)T G (i)T B (i) The method comprises the following steps:
wherein, IMF1 i For the first layer empirical mode decomposition component, corresponding to the ith value, IMF1 i+t Decomposing components for the first layer of empirical mode, wherein the i+t number is corresponding to the first layer of empirical mode;
scaling the RGB three channel signature sequence to within the range of 0-255 according to the following formula:
wherein Round () is a rounding function; k=r, G, B represent image color channels, R, G and B channels correspond to the first three-stage eigenmode components, respectively; t (T) k (i) Representing an RGB three-channel feature sequence;
and displaying the color characteristics corresponding to all points on the symmetrical lattice diagram.
6. The PMSM fault diagnosis method based on image color features according to claim 1, wherein the step of improving shape distribution and color features of the DenseNet learning image by fusing the scSE module, constructing a classification network model, extracting importance degrees on channels and spaces respectively for the input feature map, and performing addition processing, and then performing enhanced excitation on feature subgraphs with important channels and important spatial features at the same time, so as to promote network learning of more meaningful feature information.
7. The PMSM fault diagnosis method according to claim 6, wherein the improving the shape distribution and the color characteristics of the DenseNet learning image by fusing the scSE module, constructing the classification network model comprises:
input feature map p= [ P ] 1 ,p 2 ,…,p N ]Wherein the channelL and W are the height and width of the feature map, respectively;
the global spatial features are embedded into the vector v by a global pooling layer, wherein,c is the number of channels;
the obtained vector v is respectively omega by weight 1 And omega 2 Is subjected to a ReLU activation function and Sigmoid normalization processing in sequence to obtain an ith channel p i Degree of feature importance of (a)Wherein (1)>The value of (2) is
The number of the through channels is C, and the weight is omega sq The input feature map P is channel compressed by a 1×1 convolution block, the number of output channels is 1, and the feature map fq=ω with a size of lxw sq *P；
Normalizing the obtained feature map fq through Sigmod to obtain the spatial information importance degree sigma (fq (i, j)) of each spatial position (i, j) in the feature map, and enhancing the important spatial position features;
for a pair ofAnd sigma (fq (i, j)) to obtain a feature sub-graph P'.
8. A PMSM fault diagnosis system based on image color characteristics, comprising:
the original vibration signal acquisition module is used for acquiring original vibration signals of the motor to be tested under different states and various working conditions;
the empirical mode decomposition module is used for performing empirical mode decomposition on the original vibration signal and solving the intrinsic mode functions of different levels to obtain signal components of different levels;
the symmetrical lattice diagram conversion module is used for converting the original vibration signals into symmetrical lattice diagrams of three RGB channels according to signal components of different levels;
the classifying network model building module is used for improving the shape distribution and color characteristics of the DenseNet learning image through the fusion scSE module to build a classifying network model;
and the output module is used for inputting the symmetrical lattice diagrams of the RGB three channels into the classification network model and outputting fault diagnosis results.
9. An electronic device, comprising:
a memory storing at least one instruction; and a processor executing instructions stored in the memory to implement the PMSM fault diagnosis method based on image color features as defined in any one of claims 1 to 7.
10. A computer-readable storage medium storing a computer program, wherein the computer program when executed by a processor implements the PMSM fault diagnosis method based on image color characteristics according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311664830.XA CN117557865A (en) | 2023-12-06 | 2023-12-06 | PMSM fault diagnosis method and system based on image color characteristics |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311664830.XA CN117557865A (en) | 2023-12-06 | 2023-12-06 | PMSM fault diagnosis method and system based on image color characteristics |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117557865A true CN117557865A (en) | 2024-02-13 |
Family
ID=89816657
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311664830.XA Pending CN117557865A (en) | 2023-12-06 | 2023-12-06 | PMSM fault diagnosis method and system based on image color characteristics |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117557865A (en) |
-
2023
- 2023-12-06 CN CN202311664830.XA patent/CN117557865A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | An enhanced intelligent diagnosis method based on multi-sensor image fusion via improved deep learning network | |
| Rudolph et al. | Same same but differnet: Semi-supervised defect detection with normalizing flows | |
| Mundhenk et al. | Efficient saliency maps for explainable AI | |
| CN111709292B (en) | Compressor vibration fault detection method based on recursion diagram and deep convolution network | |
| CN109858352B (en) | Fault diagnosis method based on compressed sensing and improved multi-scale network | |
| Jiménez-Guarneros et al. | Diagnostic of combined mechanical and electrical faults in ASD-powered induction motor using MODWT and a lightweight 1-D CNN | |
| CN117269754B (en) | IPSM rotor demagnetizing and eccentric fault diagnosis method based on convolutional neural network operation | |
| CN111127316A (en) | Single face image super-resolution method and system based on SNGAN network | |
| CN111414988B (en) | Remote sensing image super-resolution method based on multi-scale feature self-adaptive fusion network | |
| Chang et al. | Intelligent fault diagnosis of rolling bearings using efficient and lightweight ResNet networks based on an attention mechanism (September 2022) | |
| Tian et al. | Compound Fault Diagnosis of Stator Interturn Short Circuit and Air Gap Eccentricity Based on Random Forest and XGBoost | |
| CN116337449A (en) | Sparse self-coding fault diagnosis method and system based on information fusion | |
| CN116597167B (en) | Permanent magnet synchronous motor small sample demagnetization fault diagnosis method, storage medium and system | |
| CN113530850B (en) | Centrifugal pump fault diagnosis method based on ESA and stacked capsule self-encoder | |
| CN113655778A (en) | Underwater propeller fault diagnosis system and method based on time-frequency energy | |
| Sun et al. | Research on chest abnormality detection based on improved YOLOv7 algorithm | |
| CN117557865A (en) | PMSM fault diagnosis method and system based on image color characteristics | |
| CN116610935A (en) | Mechanical fault detection method based on engine vibration signal multi-mode analysis | |
| CN114282580A (en) | Visual image-based permanent magnet driving motor demagnetization fault diagnosis model construction method and fault diagnosis method and system | |
| CN114358077A (en) | Permanent magnet synchronous motor demagnetization fault diagnosis model construction method and fault diagnosis method and system | |
| CN113673670A (en) | Power transmission line image augmentation method and system based on improved DCGAN | |
| CN116010783B (en) | RSVP weak hidden target induced electroencephalogram identification method, device and storage medium | |
| CN114383846B (en) | Bearing composite fault diagnosis method based on fault label information vector | |
| Li et al. | Research on a Small-Sample Fault Diagnosis Method for UAV Engines Based on an MSSST and ACS-BPNN Optimized Deep Convolutional Network | |
| CN117949823B (en) | Motor fault diagnosis method and device based on improved transfer learning model |
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 |