CN112304614B - Intelligent fault diagnosis method for end-to-end rolling bearing by adopting multi-attention mechanism - Google Patents
Intelligent fault diagnosis method for end-to-end rolling bearing by adopting multi-attention mechanism Download PDFInfo
- Publication number
- CN112304614B CN112304614B CN202011359124.0A CN202011359124A CN112304614B CN 112304614 B CN112304614 B CN 112304614B CN 202011359124 A CN202011359124 A CN 202011359124A CN 112304614 B CN112304614 B CN 112304614B
- Authority
- CN
- China
- Prior art keywords
- rolling bearing
- fault
- data set
- image
- fault diagnosis
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/04—Bearings
- G01M13/045—Acoustic or vibration analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
Abstract
The invention discloses an intelligent fault diagnosis method for an end-to-end rolling bearing by adopting a multi-attention mechanism. The method combines fault feature extraction and fault mode classification, and realizes weighted expression of various fault features through a multi-attention mechanism; the intelligent fault diagnosis method for the end-to-end rolling bearing by adopting the multi-attention mechanism comprises a rolling bearing data acquisition and vibration signal conversion method into an image, wherein the vibration signal conversion method comprises the steps of integrating and calculating vibration acceleration signals of the rolling bearing to obtain corresponding speeds and displacements, and combining the acceleration signals, the speed signals and the displacement signals to obtain an image with enhanced characteristics. The invention overcomes the defects that the deep neural network is easily affected by non-sensitive characteristics, so that the accuracy of the fault diagnosis of the rolling bearing is limited, and meanwhile, a great amount of time is required to generate training samples and the requirement on professional knowledge is high; the method has the advantages of high diagnosis precision, high diagnosis speed and simple operation.
Description
Technical Field
The invention relates to a fault diagnosis method for a rolling bearing, in particular to an intelligent fault diagnosis method for an end-to-end rolling bearing by adopting a multi-attention mechanism.
Background
Rolling bearings are important equipment basic components and mechanical general elements which are widely applied, are indispensable in the equipment manufacturing industry, directly determine the performance, quality and reliability of important equipment and host products, and are known as industrial joints. But at the same time, the rolling bearing is also one of the vulnerable parts of the rotary machine because the rolling bearing is often in a severe working environment and has the characteristics of high running speed, complex structure and easy failure. It is counted that more than 70% of faults of the rotating machinery are related to bearing faults, and once the bearing breaks down, series cascading faults can be caused, and the running safety of the whole equipment is seriously and directly affected. Therefore, the state monitoring and fault diagnosis of the rolling bearing have very important significance, and are always one of the important development directions in the mechanical fault diagnosis.
With the rapid development of artificial intelligence (artificial intelligence, AI), machine learning methods have found very wide application in the condition monitoring and fault diagnosis of rolling bearings. However, the conventional intelligent fault diagnosis method has problems in that manual selection of features and a large amount of tag data are required for training, and in order to solve the problems in fault diagnosis, the emerging deep learning method is gradually applied to the field of fault diagnosis by people. In 2006, hinton et al proposed using a self-encoder (autoencoder) to reduce the dimensionality of the data and a pre-training approach to quickly train a deep belief network to suppress the gradient vanishing problem. With the method as a mark, deep learning is taken as a new method in the emerging pattern recognition field, and breakthrough progress is made in aspects of image recognition, voice recognition, natural language processing and the like. Meanwhile, due to the deep learning multilayer structure, deep relation can be extracted from a large amount of data, and great attention and application are also obtained in the field of bearing fault diagnosis. The existing rolling bearing fault diagnosis usually adopts a deep neural network, and the deep neural network is easily influenced by non-sensitive characteristics, so that the accuracy of the rolling bearing fault diagnosis is influenced.
During processing of large amounts of data, data of different characteristics requires different methods to properly display the characteristics. In order to realize fault diagnosis of the rolling bearing, many scholars combine vibration signals and deep learning technology, and a time-frequency chart, a histogram and other methods are used for converting the vibration signals into images for classification, but a great deal of time is required for generating training samples, and the training samples are greatly dependent on expert experience knowledge.
Therefore, there is a need to develop a method for intelligently and rapidly diagnosing the failure of the rolling bearing.
Disclosure of Invention
The invention aims to provide an intelligent fault diagnosis method of an end-to-end rolling bearing by adopting a multi-attention mechanism, which is a fault diagnosis method of the rolling bearing based on the combination of CBAM-ResNet, and utilizes the characteristics of corresponding speed and displacement obtained by integral transformation of vibration acceleration signals, the speed signals and the displacement signals are combined to obtain an image with enhanced characteristics, then a channel attention and space attention structure (Convolutional Block Attention Module, CBAM) is introduced into a deep residual network to realize the feature extraction of the vibration signals, finally the fault diagnosis of the rolling bearing is realized by utilizing a multi-classification function (wherein the multi-classification function refers to the output classification function of a fault diagnosis model (CBAM-ResNet)), the distribution characteristics of the fault mode can be captured, the diagnosis precision is high, the diagnosis speed is high, and the operation is simple.
In order to achieve the above purpose, the technical scheme of the invention is as follows: the intelligent fault diagnosis method for the end-to-end rolling bearing by adopting the multi-attention mechanism is characterized by comprising the following steps of: the fault feature extraction and the fault mode classification are combined, and the weighted expression of various fault features is realized through a multi-attention mechanism;
the intelligent fault diagnosis method for the end-to-end rolling bearing by adopting the multi-attention mechanism comprises a rolling bearing data acquisition and vibration signal conversion method into an image, wherein the vibration signal conversion method comprises the steps of integrating and calculating vibration acceleration signals of the rolling bearing to obtain corresponding speeds and displacements, and combining the acceleration signals, the speed signals and the displacement signals to obtain an image with enhanced characteristics.
In the technical proposal, the intelligent fault diagnosis method of the end-to-end rolling bearing adopting the multi-attention mechanism comprises the following steps which are executed in turn,
step one: collecting rolling bearing data;
step two: preprocessing rolling bearing data;
the rolling bearing data preprocessing is to convert vibration signals of the rolling bearing into images; the method for converting the vibration signal of the rolling bearing into the image comprises the following steps: the vibration acceleration signal of the rolling bearing is subjected to integral calculation to obtain corresponding speed and displacement, and the acceleration signal, the speed signal and the displacement signal are combined to obtain an image with enhanced characteristics;
step three: constructing a CBAM-ResNet diagnosis model and analyzing and constructing diagnosis results;
and (3) sending the image in the step one into a CBAM-ResNet diagnostic model for training, and classifying a test data set by using the trained CBAM-ResNet diagnostic model.
In the above technical solution, the multi-attention mechanism adopted is a channel attention and spatial attention structure.
In the above technical solution, in the second step, the method for converting the vibration signal into the image specifically includes the following steps,
according to four categories of outer ring faults, inner ring faults, rolling body faults and normal conditions of the rolling bearing, according to the standard of 50% of the overlapped signals, intercepting vibration signal samples to obtain a vibration acceleration signal data set { AC } i I=1, 2, …, m×m×l }, where L represents the total number of samples and m×m represents the pixel size of the image;
obtained by integration, velocity signal dataset { VE i I=1, 2, …, mxmx×l }, displacement signal dataset { DIS } i |i＝1,2,…,M×M×L}；
Then, the vibration acceleration signal data set, the velocity signal data set and the displacement signal data set are substituted into (6) respectively for processing so that the data range is converted between [0,255],
in the formula (6), P (m, m), m=1, 2 … j represents the pixel point of the image, and the function round (·) is a rounding function, x i Represents the ith sample, x of the dataset min Representing the minimum value of the samples in the dataset, x max Representing a maximum value of samples in the dataset;
in the generated RGB image, the red channel pixel values are filled with the acceleration signal data set, the green channel pixel values are filled with the velocity signal data set, and the blue channel pixel values are filled with the displacement signal data set.
In the above technical solution, in step three, in the original res net network structure, a CBAM attention module is added after each group of residual blocks, and the output number of the output layer is set to 4, so as to match with four health types of an outer ring fault, an inner ring fault, a rolling body fault and a normal condition of the rolling bearing.
The invention has the following advantages:
(1) The invention builds the deep network by introducing the residual block, thus solving the problems of overfitting, gradient disappearance or gradient explosion of the deep neural network model in the training process; in the invention, in the original ResNet network structure, a CBAM attention module is added after each group of residual blocks, and the feature extraction capacity of the model is improved by introducing the CBAM attention module;
(2) The invention uses the attention mechanism to selectively characterize, thereby effectively overcoming the problem that the deep neural network is easily affected by non-sensitive characteristics, and more fully utilizing the characteristics and the information among the characteristics;
(3) According to the invention, fault feature extraction and fault mode classification are fused together, and the attention mechanism can realize weighted expression of different features, so that the classified features have more expression capability;
(4) The method utilizes the characteristic that the vibration acceleration signal can obtain corresponding speed and displacement through integral calculation, combines the acceleration signal, the speed signal and the displacement signal to obtain the image with the enhanced characteristic, and can rapidly obtain the image with the enhanced characteristic from the original data without presetting parameters or expert experience; and then the graph is used as a model to be input into a CBAM-ResNet diagnosis model for training and diagnosis, so that the distribution characteristics of a fault mode can be captured, and the diagnosis speed and the diagnosis precision are improved.
(5) The classification precision of the model training set of the invention is close to 100%, and the fault diagnosis precision of the rolling bearing reaches 98.33%.
Drawings
Fig. 1 is a residual unit diagram of a deep residual network in the present invention.
Fig. 2 is a block diagram of a convolutional neural network incorporating CBAM in accordance with the present invention.
Fig. 3 is a diagnostic flow chart of the present invention.
Fig. 4 is a flowchart of a method for converting a vibration signal into an image according to the present invention.
FIG. 5 is a block diagram of the CBAM-ResNet model of the present invention.
Fig. 6 is a structural diagram of a laboratory bench used in example 1 of the present invention.
Fig. 7 is a flowchart of data conversion into an image in embodiment 1 of the present invention.
FIG. 8 is a graph of accuracy and loss function values versus iteration number for the CBAM-ResNet model in example 1 of the present invention in a training set.
Fig. 9 is a confusion matrix diagram of the diagnostic result in example 1 of the present invention.
Fig. 10 is a structural diagram of a laboratory bench used in example 2 of the present invention.
Fig. 11 is a time domain diagram of vibration signals for each health type in embodiment 2 of the present invention.
FIG. 12 is a graph of accuracy and loss function values versus iteration number for the CBAM-ResNet model in example 2 of the present invention in a training set.
Fig. 13 is a confusion matrix diagram of the diagnostic result in example 2 of the present invention.
Detailed Description
The following detailed description of the invention is, therefore, not to be taken in a limiting sense, but is made merely by way of example. While making the advantages of the present invention clearer and more readily understood by way of illustration.
As can be seen with reference to the accompanying drawings: the intelligent fault diagnosis method of the end-to-end rolling bearing adopts a multi-attention mechanism, combines fault feature extraction and fault mode classification, realizes the weighted expression of various fault features through the multi-attention mechanism, and can selectively characterize the fault features by using the attention mechanism, thereby effectively overcoming the problem that a deep neural network is easily affected by insensitive features and more fully utilizing the features and information among the features; because the attention mechanism has the excellent characteristics, the invention focuses on important characteristics and simultaneously suppresses unnecessary characteristics in the characteristic extraction stage, improves the characteristic expression capacity of the convolutional neural network model, and further improves the accuracy of the fault diagnosis of the rolling bearing on the premise of not obviously increasing the calculated amount and the parameter; the invention is characterized in that the collected vibration signals are directly used for completing fault mode classification, and a manual characteristic extraction process is not needed in the middle;
the intelligent fault diagnosis method for the end-to-end rolling bearing by adopting the multi-attention mechanism comprises the steps of collecting rolling bearing data and converting vibration signals into images, wherein the method for converting the vibration signals into the images is that vibration acceleration signals of the rolling bearing are subjected to integral calculation to obtain corresponding speeds and displacements, then the acceleration signals, the speed signals and the displacement signals are combined to obtain images with enhanced characteristics, and the fault diagnosis of the end-to-end rolling bearing is rapidly and accurately realized through a simple diagnosis flow.
Further, the intelligent fault diagnosis method for the end-to-end rolling bearing by adopting the multi-attention mechanism comprises the following steps of sequentially executing,
step one: collecting rolling bearing data; wherein, the original vibration signal is collected as the prior art;
step two: preprocessing rolling bearing data;
the rolling bearing data preprocessing is to convert vibration signals of the rolling bearing into images (as shown in fig. 4, fig. 4 is sample segmentation and forms images by combining acceleration, speed and displacement signals) as model input; the method for converting the vibration signal of the rolling bearing into the image comprises the following steps: the vibration acceleration signal of the rolling bearing is subjected to integral calculation to obtain corresponding speed and displacement, and the acceleration signal, the speed signal and the displacement signal are combined to obtain an image with enhanced characteristics; data preprocessing is the first step of deep learning and is also an important step; in the process of processing a large amount of data, the data with different characteristics can be accurately displayed by different methods, and the method for converting the vibration signal of the rolling bearing into the image can quickly obtain the image with the enhanced characteristic from the original data without presetting parameters or expert experience, so that the method has the characteristics of rapidness and intelligence, and can rapidly and accurately realize end-to-end fault diagnosis of the rolling bearing through a simple diagnosis flow; the defects that in the prior art, in order to realize fault diagnosis of the rolling bearing, vibration signals are converted into images for classification by using methods such as time-frequency diagrams, histograms and the like in combination with vibration signals and deep learning technology, a great deal of time is required for generating training samples, expert experience knowledge is greatly relied on, time is consumed, and requirements on expert knowledge are high are overcome;
step three: constructing a CBAM-ResNet diagnosis model and analyzing and constructing diagnosis results;
the image in the first step is sent to a CBAM-ResNet diagnosis model for training, and a test data set (shown in figure 3) is classified by using the trained CBAM-ResNet diagnosis model, wherein the test data set is the image preprocessed in the second step; step two, preprocessing the image to form a data set, and dividing the data set into a training data set and a testing data set; the invention does not need to manually participate in fault feature extraction and fault mode classification, and the fault feature extraction and the fault mode classification are completed by a CBAM-ResNet diagnosis model, thereby having the characteristics of rapidness and intelligence.
Both the CBAM model and the res net model are prior art.
Further, the multi-attention mechanism employed is a channel attention and spatial attention structure (as shown in fig. 2); the CBAM model is a model combining channel attention and spatial attention structure; channel attention and spatial attention structure (CBAM) uses the attention mechanism in the channel dimension and the spatial dimension, respectively, emphasizes significant features in both the space and channel dimensions, focuses on important features and suppresses unnecessary features; meanwhile, the channel attention structure and the space attention structure are mutually independent, so that the channel attention structure and the space attention structure can be used as independent modules in the existing convolutional neural network architecture;
the process of the multi-attention mechanism (CBAM) includes the following two operations:
in the formulas (4) and (5),
representing the characteristics of the input, C x H x B representing the dimensions of each channel,
represents the attention weight in the channel dimension, +.>
Representing the attention weight in the spatial dimension, the symbol +.>
Representing element-by-element multiplication; r represents a feature space; c×1×1 represents a channel dimension; 1 XH B represents the spatial dimension; f' represents a weighted value obtained by multiplying the input characteristics element by element and the attention weight of the channel; f "represents the weighted value of F' after multiplying the spatial attention weight element by element.
Further, in the second step, the method for converting the vibration signal into the image specifically includes the following steps,
according to four categories of outer ring faults, inner ring faults, rolling body faults and normal conditions of the rolling bearing, according to the standard of 50% of overlapped signals, cutting vibration signal samples (for example, starting from the 1 st point of vibration signal data, cutting the nth point as a first signal sample, then starting from the (N-m) th point of next signal sample, cutting the (2N-m) th point as a second signal sample, and so on to realize segment overlapping cutting of signals, and totally cutting N signal samples, wherein N is a positive integer greater than or equal to 1, m is less than N, and N is a positive integer), so as to obtain vibration acceleration signal dataSet { AC i I=1, 2, …, m×m×l }, where L represents the total number of samples and m×m represents the pixel size of the image; i represents the sequence number of each data in the data set;
obtained by integration, velocity signal dataset { VE i I=1, 2, …, mxmx×l }, displacement signal dataset { DIS } i I=1, 2, …, m×m×l }; wherein i represents the number of sequences of each data in the dataset; m×m represents the pixel size of an image; l represents the total number of samples;
the method for obtaining corresponding speed and displacement by integrating the vibration acceleration signals of the rolling bearing is the prior art;
then, the vibration acceleration signal data set, the speed signal data set and the displacement signal data set are respectively substituted into formula (6) to be processed, so that the data range is transformed into [0,255] (wherein, data normalization is carried out for conversion into RGB images; the RGB images require the value range of each pixel point to be [0,255 ]), and RGB images are generated;
wherein, the formula (6) is used for normalizing the value of each pixel point on three channels of the RGB image to [0,255];
in the formula (6), P (m, m), m=1, 2 … j represents a pixel point of an image, and j is a positive integer; p (m, m) represents each pixel point on the image; the function round (·) is a rounding function; x is x i Representing the ith sample of the dataset; x is x min Representing a sample minimum in the dataset; x is x max Representing a maximum value of samples in the dataset;
in the generated RGB image, the red channel pixel values are filled with the acceleration signal data set, the green channel pixel values are filled with the velocity signal data set, and the blue channel pixel values are filled with the displacement signal data set (as shown in fig. 4).
Further, in step three, in the original res net network structure, a CBAM attention module is added after each group of residual blocks, and the output number of the output layer is set to be 4, so as to match the four health types of the outer ring fault, the inner ring fault, the rolling body fault and the normal condition of the rolling bearing (as shown in fig. 5);
the original ResNet network consists of a series of residual blocks, the residual block structure is shown in figure 1, when the input is x, the learned characteristic is marked as H (x), the residual of the structure is F (x) =H (x) -x, when the residual F (x) =0, the residual is only mapped with identity, the deep network performance is not reduced, and when the residual F (x) notequal to 0, the residual layer learns new characteristic on the basis of the input characteristic, so that the deep network has better performance;
let x be l And x l+1 Respectively representing the input and output of the layer I residual block, F (x l ,W l ) Is a residual function representing the learned residual, W l Is a weight vector to be learned, one residual block can be expressed as:
y l ＝h(x l )+F(x l ,W l ) (1)
x l+1 ＝f(y l ) (2)
wherein f (·) is an activation function, typically using the Relu function;
in the formula (1), h (·) represents a learned feature;
in formula (2), f (·) is a Relu activation function; y is l Representing the output of the residual block before activating the function;
based on the formulas (1) and (2), the learning characteristics from the shallow layer L to the deep layer L are obtained as follows:
wherein x is L Representing the output of the layer L residual block, W i Is a weight vector to be learned;
in the formula (3), x l An output representing an L-layer residual block; x is x l Representing an output representing a layer I residual block; f (x) i ，w i ) Representing a residual function; x is x i Representing the input of the i-layer residual block.
Examples
The invention will now be described in detail with reference to the embodiment of the invention applied to the diagnosis of the faults of rolling bearings at the driving end of a certain motor, and the invention has guiding function for the fault diagnosis of other rolling bearings.
Example 1
In this embodiment, a published rolling bearing dataset from the CWRU (kesixi Chu Da, case Western Reserve University) bearing data center is used.
The fault diagnosis method for a certain rolling bearing in the embodiment comprises the following steps,
s1: collecting data;
the experiment table adopted in the embodiment is shown in fig. 6, and the bearing experiment table mainly comprises a motor, a sensor, a rolling bearing and a dynamometer; the rolling bearing vibration data of the motor driving end and the fan end are obtained by an acceleration sensor arranged on the induction motor shell, and the sampling frequency is 12kHz;
besides the normal state, the rolling bearing also artificially introduces three fault states of an inner ring fault, an outer ring fault and a rolling body fault in an electric spark machining mode; in addition, each fault class has three defect diameters (0.007,0.014 and 0.021 inches), and bearing vibration signals are acquired at three rotating speeds respectively; in this example, a study was conducted on the classification of rolling bearing faults at the motor drive end of 1797rpm (load 0) and a defect diameter of 0.021 inches; therefore, 4 health types including normal state in the sample of the embodiment, namely, inner ring failure, outer ring failure, rolling body failure, and normal state;
according to four categories of outer ring faults, inner ring faults, rolling body faults and normal conditions of the rolling bearing, the rolling bearing data acquisition method is adopted to obtain a vibration acceleration signal data set { AC } i |i＝1,2,…,M×M×L}；
Obtained by integration, velocity signal dataset { VE i I=1, 2, …, mxmx×l }, displacement signal dataset { DIS } i |i＝1,2,…,M×M×L}；
S2: converting the data into an image;
the vibration data of four health types are converted into RGB images by adopting the method for converting the vibration signals of the rolling bearing into images, and the result is shown in figure 7; wherein each health type has 600 RGB images with pixel size of 32×32, and contains 2400 pieces; randomly selecting 95% of data in the samples as training data, and the rest 5% of sample data as test samples, namely 2280 samples for training and 120 samples for testing;
as can be seen from the converted image (as shown in fig. 7), the rolling bearing is distinguished from RGB images of different health types; the image pigment points of the outer ring faults are the most dense, compared with the image pigment points of the inner ring faults are reduced, the image pigment points of the rolling body faults are the sparsest, and the images of the normal type have obvious stripe characteristics;
s3: analyzing the diagnosis result of the CBAM-ResNet model;
after the conversion from data to images, the data set is sent to a CBAM-CNN model for training, and then the trained model is used for classifying the test data set; the training and the testing of the model are carried out on the same workstation, namely, the model is realized by using Python3.6 programming under a TensorFlow platform in a GeForce RTX 2080Ti video card (11 GB video memory) under a windows10 environment;
in the training process, the iteration times are set to 120, the batch size is set to 32, the Adam optimization algorithm is adopted to update network parameters, the initial learning rate is set to 0.001, and the learning rate is dynamically changed in the training process; the relation diagram of the accuracy and the loss function value of the CBAM-ResNet model in the training set and the iteration times is shown in figure 8;
as can be seen from fig. 8, the model training results are: the classification precision of the training set is close to 100%, which shows that the model fitting effect is good, the loss function is smoothly reduced and tends to be stable, and the loss function is not sunk into local optimum; then, inputting the data of the test set into a trained diagnosis model, and taking an average value of the results through 10 times of random experiments to obtain the classification accuracy of the test set reaching 98.33%, which shows that the diagnosis method can capture the distribution characteristics of the fault mode. To further understand the detailed classification of each health type, a confusion matrix of diagnostic results is plotted, as shown in fig. 9; as can be seen from fig. 9, in the test set, one inner ring failure is misjudged as a rolling element failure, one outer ring failure is misjudged as an inner ring failure, and all normal cases are correctly classified.
The following experiments of CWRU bearing fault data sets further illustrate the characteristic that the vibration acceleration signals can be used for obtaining corresponding speed and displacement through integral calculation, and the three acceleration signals, the speed signals and the displacement signals are combined to form the effectiveness and the superiority of the RGB image with enhanced characteristics.
The acceleration signal, the speed signal and the displacement signal are respectively and independently formed into an RGB image, the acceleration signal and the speed signal are formed into an RGB image, the acceleration signal and the displacement signal are formed into an RGB image, 6 conditions of the acceleration signal and the displacement signal are formed into an RGB image, and the speed signal and the displacement signal are subjected to comparison analysis with the embodiment 1 (the number 7 in the table 1) of the invention, and the obtained image data sets are respectively sent into a CBAM-ResNet rolling bearing fault diagnosis model for training and diagnosis, wherein the fault diagnosis precision of the rolling bearing is shown in the table 1.
Table 1 fault diagnosis accuracy of Rolling bearing under image dataset composed of different signals
As can be seen from table 1, when the acceleration signal, the velocity signal and the displacement signal are used alone to form an RGB image, the diagnostic accuracy of the image formed by the acceleration signal in the embodiment 1 of the present invention for performing fault diagnosis is significantly higher than that of the image formed by the velocity signal and the displacement signal; and when the acceleration signals are respectively combined with other two signals, the diagnosis precision is improved to 94.11% and 94.21% respectively. But are lower than the diagnostic accuracy of the signal conversion to image method proposed in example 1 of the present invention.
Verification method
In order to further illustrate the performance superiority of the CBAM-ResNet model in the present invention, three common classification models of SVM (support vector machine (Support Vector Machine, SVM)), BP (artificial neural network model) and CNN (convolutional neural network model) are selected and compared with the above-mentioned example 1. The SVM model uses 8 characteristic parameters of rolling bearing (which is a ball bearing) vibration signal standard deviation, kurtosis, average value, root mean square, waveform factor, peak factor, margin factor and kurtosis factor. The BP neural network also uses the time domain characteristic parameters as input, and comprises 2 hidden layers, and each layer comprises 50 neurons. And randomly selecting 95% of data in the samples as training data, and the rest 5% of sample data as test samples, wherein the results are averaged through 10 random experiments on several classification models. The fault diagnosis results are shown in table 2.
TABLE 2 results of comparative experiments on different models
As can be seen from table 2: after training by using the characteristic samples, the BP neural network has the recognition rate of 83.33%, the accumulated total number of incorrectly recognized samples is 20, the SVM bearing fault diagnosis method is used, through continuous attempts, the recognition rate can reach 91.67% after the relatively proper parameters are selected empirically, and the classification result of the CNN model can reach 95.83%, but is still lower than the classification result of the CBAM-ResNet model in the embodiment 1 of the invention.
Example 2
The present embodiment employs a local laboratory bearing failure dataset.
In the embodiment, a mechanical fault comprehensive simulation experiment table of SpectraQuest company is used for carrying out a rolling bearing fault diagnosis test, and the specific diagnosis method comprises the following steps,
s1: data acquisition and pretreatment;
as shown in fig. 10, the experiment table in the embodiment mainly comprises a rolling bearing, a detachable bearing seat, a motor, a speed regulating device and the like; the experiment of the embodiment is carried out on an ER-16K rolling bearing matched with the experiment table, pitting faults are prefabricated on an inner ring, an outer ring and a rolling body by utilizing electric sparks, a piezoelectric acceleration sensor arranged above a bearing seat is used for picking up vibration signals of the rolling bearing in the experiment, the sampling frequency is 12kHz, and after amplification and filtration, the signals are acquired by a data acquisition system;
the embodiment classifies four health types of an inner ring fault, an outer ring fault and a rolling body fault of the motor side rolling bearing including a normal state;
the collecting method of this embodiment is the same as that of embodiment 1;
after enough data are obtained, the vibration acceleration signals are utilized to obtain corresponding speed and displacement signals through integral calculation, and a vibration signal time domain diagram of each health type is shown in fig. 11; as can be seen from fig. 11, the amplitudes of the three types of signals for each health type are greatly different;
s2: converting the data into an image;
the method of converting the data into an image is the same as that of example 1, and four health-type RGB images are obtained; the size and arrangement of the data set is the same as in example 1;
s3: analyzing a diagnosis result;
the parameter settings of the model training process were the same as in example 1; the relation diagram of the accuracy and the loss function value of the CBAM-ResNet model in the training set and the iteration times in the embodiment is shown in fig. 12; as can be seen from fig. 12, by training the diagnostic model, as the number of iterations increases, the training set classification accuracy gradually increases and tends to be stable, and at the same time, the loss function smoothly decreases and tends to be stable;
then, inputting the data of the test set into a trained diagnosis model, and taking an average value of results through 10 random experiments to obtain the classification accuracy of the test set of 97.50%; to further understand the detailed classification of each health type, a confusion matrix of diagnostic results is plotted, as shown in fig. 13; as can be seen from fig. 13, in the test set, there were 3 classifications of rolling element faults, and all of the remaining three health types were correctly classified.
The invention further illustrates the characteristic that the vibration acceleration signals can be used for obtaining the corresponding speed and displacement through integral calculation through experiments of a local laboratory bearing fault data set, and the three signals of the acceleration, the speed and the displacement are combined to form the RGB image with enhanced characteristics, namely the effectiveness, the superiority and the generalization performance.
The acceleration signal, the speed signal and the displacement signal are respectively and independently formed into an RGB image, the acceleration signal and the speed signal are formed into an RGB image, the acceleration signal and the displacement signal are formed into an RGB image, 6 conditions of the acceleration signal and the displacement signal are formed into an RGB image, and the speed signal and the displacement signal are subjected to comparison analysis with the image of the embodiment 2 (the number 7 in the table 3), and the obtained image data sets are respectively sent into a CBAM-ResNet rolling bearing fault diagnosis model for training and diagnosis, wherein the fault diagnosis precision of the rolling bearing is shown in the table 3.
TABLE 3 Rolling bearing fault diagnosis precision under different signal composition image datasets
As can be seen from table 3, the diagnostic accuracy of the fault diagnosis using the image composed of the acceleration signal is significantly higher than that of the fault diagnosis using the image composed of the velocity signal and the displacement signal; the diagnostic accuracy of fault diagnosis of the image formed by combining the acceleration signal with the speed signal and the displacement signal can reach 94.70% and 94.18% respectively, but is lower than that of embodiment 2 of the invention.
Verification method
In order to further illustrate the performance superiority of the CBAM-ResNet model in the invention, three common fault diagnosis models of SVM, BP and CNN are selected in the embodiment to carry out comparison analysis with the embodiment 2 of the invention, and the parameter setting is the same as that of the embodiment 1; taking an average value of the results through 10 random experiments; the fault diagnosis results are shown in table 4.
TABLE 4 results of comparative experiments on different models
As can be seen from Table 4, the classification accuracy obtained by the CBAM-ResNet diagnostic model in the embodiment 2 of the invention is improved by 15.83% compared with the BP neural network, the fault diagnosis accuracy of the rolling bearing is obviously improved, and the fact that the CBAM-ResNet diagnostic model in the invention can more accurately capture the hidden characteristics of the data set is also verified.
The embodiment of the invention demonstrates the effectiveness and generalization capability of the invention through the verification and analysis of two different embodiments; meanwhile, the invention omits a complicated artificial feature extraction process in the diagnosis process, and realizes end-to-end fault diagnosis while reducing the miss rate of the rolling bearing fault.
The invention utilizes the characteristic that the convolutional neural network can enhance the nonlinear characterization capability of the fault diagnosis model, introduces a channel attention mechanism and a space attention mechanism on the basis of the characteristic, and models the nonlinear relation between the characteristics; the experimental results of the two embodiments show that: (1) The characteristics of corresponding speed and displacement are obtained by integrating the fault vibration acceleration signals of the rolling bearing, and the image with enhanced characteristics is obtained after the acceleration signals, the speed signals and the displacement signals are combined, so that the fault diagnosis of the rolling bearing can be used; (2) The established CBAM-ResNet diagnosis model can automatically extract characteristics and complete end-to-end fault diagnosis of the rolling bearing; (3) Compared with the existing data driving fault diagnosis method, the method of the invention has better diagnosis precision and better robustness in the rolling bearing fault diagnosis.
Other non-illustrated parts are known in the art.
Claims (3)
1. The intelligent fault diagnosis method for the end-to-end rolling bearing by adopting the multi-attention mechanism is characterized by comprising the following steps of: the fault feature extraction and the fault mode classification are combined, and the weighted expression of various fault features is realized through a multi-attention mechanism; selectively characterizing by using an attention mechanism, focusing on important features and simultaneously inhibiting unnecessary features in a feature extraction stage, and directly using collected vibration signals to finish fault mode classification;
the specific method comprises the following steps sequentially executed,
step one: collecting rolling bearing data;
step two: preprocessing rolling bearing data;
the rolling bearing data preprocessing is to convert vibration signals of the rolling bearing into images; the method for converting the vibration signal of the rolling bearing into the image comprises the following steps: the vibration acceleration signal of the rolling bearing is subjected to integral calculation to obtain corresponding speed and displacement, and the acceleration signal, the speed signal and the displacement signal are combined to obtain an image with enhanced characteristics;
specifically comprises the following steps of,
according to four categories of outer ring faults, inner ring faults, rolling body faults and normal conditions of the rolling bearing, according to the standard of 50% of the overlapped signals, vibration signal samples are intercepted, and a vibration acceleration signal data set { AC i|i=1, 2,.. The number of M multiplied by L } is obtained, wherein L represents the total number of samples, and M multiplied by M represents the pixel size of an image;
the velocity signal data set { VE i i=1, 2, & gt, m×m×l }, and the displacement signal data set { DIS i i=1, 2, & gt, m×m×l };
next, the vibration acceleration signal data set, the velocity signal data set and the displacement signal data set are substituted into (6) respectively for processing so that the data range is converted to between [0,255],
in the formula (6), P (m, m), m=1, 2..j represents a pixel point of an image, a function round (·) is a rounding function, x i represents an i-th sample of the data set, x min represents a sample minimum value in the data set, and x max represents a sample maximum value in the data set;
in the generated RGB image, red channel pixel values are filled with the acceleration signal data set, green channel pixel values are filled with the velocity signal data set, and blue channel pixel values are filled with the displacement signal data set;
step three: constructing a CBAM-ResNet diagnosis model and analyzing and constructing diagnosis results;
and (3) sending the image in the step two into a CBAM-ResNet diagnostic model for training, and classifying the test data set by using the trained CBAM-ResNet diagnostic model.
2. The intelligent fault diagnosis method for the end-to-end rolling bearing adopting the multi-attention mechanism according to claim 1, wherein the intelligent fault diagnosis method is characterized by comprising the following steps of: the multi-attentiveness mechanisms employed are channel attentiveness and spatial attentiveness structures.
3. The intelligent fault diagnosis method for the end-to-end rolling bearing adopting the multi-attention mechanism according to claim 2, wherein the intelligent fault diagnosis method is characterized by comprising the following steps of: in the third step, in the original ResNet network structure, a CBAM attention module is added after each group of residual blocks, and the output number of the output layer is set to be 4 so as to match the four health types of the outer ring fault, the inner ring fault, the rolling body fault and the normal condition of the rolling bearing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011359124.0A CN112304614B (en) | 2020-11-27 | 2020-11-27 | Intelligent fault diagnosis method for end-to-end rolling bearing by adopting multi-attention mechanism |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011359124.0A CN112304614B (en) | 2020-11-27 | 2020-11-27 | Intelligent fault diagnosis method for end-to-end rolling bearing by adopting multi-attention mechanism |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112304614A CN112304614A (en) | 2021-02-02 |
| CN112304614B true CN112304614B (en) | 2023-04-25 |
Family
ID=74486967
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011359124.0A Active CN112304614B (en) | 2020-11-27 | 2020-11-27 | Intelligent fault diagnosis method for end-to-end rolling bearing by adopting multi-attention mechanism |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112304614B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113340598A (en) * | 2021-06-01 | 2021-09-03 | 西安交通大学 | Rolling bearing intelligent fault diagnosis method based on regularization sparse model |
| CN113449618A (en) * | 2021-06-17 | 2021-09-28 | 南京航空航天大学 | Method for carrying out deep learning rolling bearing fault diagnosis based on feature fusion and mixed enhancement |
| CN114118140B (en) * | 2021-10-29 | 2022-09-23 | 新黎明科技股份有限公司 | Multi-view intelligent fault diagnosis method and system for explosion-proof motor bearing |
| CN114563186B (en) * | 2022-01-21 | 2023-01-06 | 苏州大学 | Rolling bearing unbalance fault diagnosis method and system |
| CN114993677B (en) * | 2022-05-11 | 2023-05-02 | 山东大学 | Rolling bearing fault diagnosis method and system for unbalanced small sample data |
| CN114992138B (en) * | 2022-06-13 | 2024-03-26 | 湘潭大学 | GAN-CNN-BiLSTM-based industrial circulating water system water supply pump fault diagnosis method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN209085657U (en) * | 2017-08-02 | 2019-07-09 | 强力物联网投资组合2016有限公司 | For data gathering system related or industrial environment with chemical production technology |
| CN111833307A (en) * | 2020-06-15 | 2020-10-27 | 湖南科技大学 | Method and device for diagnosing cracks of wind turbine blade |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108334936B (en) * | 2018-01-30 | 2019-12-24 | 华中科技大学 | Fault prediction method based on migration convolutional neural network |
| CN109765053B (en) * | 2019-01-22 | 2022-04-29 | 中国人民解放军海军工程大学 | Rolling bearing fault diagnosis method using convolutional neural network and kurtosis index |
-
2020
- 2020-11-27 CN CN202011359124.0A patent/CN112304614B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN209085657U (en) * | 2017-08-02 | 2019-07-09 | 强力物联网投资组合2016有限公司 | For data gathering system related or industrial environment with chemical production technology |
| CN111833307A (en) * | 2020-06-15 | 2020-10-27 | 湖南科技大学 | Method and device for diagnosing cracks of wind turbine blade |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112304614A (en) | 2021-02-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112304614B (en) | Intelligent fault diagnosis method for end-to-end rolling bearing by adopting multi-attention mechanism | |
| CN109765053B (en) | Rolling bearing fault diagnosis method using convolutional neural network and kurtosis index | |
| CN109186973B (en) | Mechanical fault diagnosis method of unsupervised deep learning network | |
| CN109858352B (en) | Fault diagnosis method based on compressed sensing and improved multi-scale network | |
| CN110516305B (en) | Intelligent fault diagnosis method under small sample based on attention mechanism meta-learning model | |
| CN105841961A (en) | Bearing fault diagnosis method based on Morlet wavelet transformation and convolutional neural network | |
| CN112257530B (en) | Rolling bearing fault diagnosis method based on blind signal separation and support vector machine | |
| CN108510153A (en) | A kind of multi-state rotary machinery fault diagnosis method | |
| CN110702411A (en) | Residual error network rolling bearing fault diagnosis method based on time-frequency analysis | |
| CN113128338B (en) | Intelligent diagnosis method for printer roller faults under small sample | |
| CN113865868B (en) | Rolling bearing fault diagnosis method based on time-frequency domain expression | |
| CN112378660A (en) | Intelligent fault diagnosis method for aero-engine bearing based on data driving | |
| CN113158984B (en) | Bearing fault diagnosis method based on complex Morlet wavelet and lightweight convolution network | |
| CN111707472B (en) | Rolling bearing fault diagnosis method based on FFT and full-connection layer-SVM | |
| CN114491823B (en) | Train bearing fault diagnosis method based on improved generation countermeasure network | |
| CN112231971A (en) | Blast furnace fault diagnosis method based on relative integral trend diffusion fault sample generation | |
| CN111310719A (en) | Unknown radiation source individual identification and detection method | |
| CN114997211A (en) | Cross-working-condition fault diagnosis method based on improved countermeasure network and attention mechanism | |
| CN115290326A (en) | Rolling bearing fault intelligent diagnosis method | |
| Chen et al. | An adversarial learning framework for zero-shot fault recognition of mechanical systems | |
| CN113916535A (en) | Bearing diagnosis method, system, equipment and medium based on time frequency and CNN | |
| CN111783941B (en) | Mechanical equipment diagnosis classification method based on probability confidence convolutional neural network | |
| CN115905976B (en) | Attention mechanism-based fault diagnosis method, system and equipment for Highway Bi-LSTM bearing | |
| CN112766200A (en) | Planetary gear box small sample fault diagnosis method based on one-dimensional convolutional neural network | |
| CN110490218B (en) | Rolling bearing fault self-learning method based on two-stage DBN |
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 |