CN113946920A - Rolling bearing fault diagnosis method with unbalanced data and data set deviation - Google Patents

Abstract

The invention provides a rolling bearing fault diagnosis method with unbalanced data and deviated data set. The invention comprises a generator G, a discriminator D and a cross-domain fault classifier C, the whole fault diagnosis model is alternately confronted and trained, the purpose is to use the synthesized fault sample to rebalance the training set data, and use the rebalance training set data and the unlabelled test set sample to train the fault classifier C. And performing fault diagnosis on the test set fault sample by using the trained cross-domain fault middle classifier C. The method has high fault diagnosis accuracy, and is very suitable for fault diagnosis of the rolling bearing with unbalanced fault data and deviation of a data set.

Description

Rolling bearing fault diagnosis method with unbalanced data and data set deviation

Technical Field

The invention relates to the technical field of rolling bearing fault processing, in particular to a rolling bearing fault diagnosis method with unbalanced data and deviation of a data set.

Background

In industrial application, the rolling bearing is in a normal state most of the time, the fault probability is low, and the operation condition of the rolling bearing is changed frequently. The collected rolling bearing fault samples have two problems that (1) fault data are unbalanced and (2) data set deviation exists, and the accuracy of the traditional intelligent fault diagnosis method is seriously reduced.

Disclosure of Invention

In view of the above-mentioned technical problems, a rolling bearing fault diagnosis method is provided in which data is unbalanced and data sets are deviated. The invention is used for fault diagnosis of a rolling bearing with unbalanced fault data and data set deviation, and the principle is that low-proportion fault samples without data set deviation are automatically generated, and the low-proportion fault samples without data set deviation balance fault data for fault model training. The technical means adopted by the invention are as follows:

a rolling bearing fault diagnosis method with unbalanced data and deviated data set comprises the following steps:

step 1, collecting vibration signals of a rolling bearing as fault data, dividing the fault data into a training set and a testing set, wherein a fault sample of the training set and a fault sample of the testing set have deviation;

step 2, establishing a fault diagnosis model, wherein the fault diagnosis model comprises a rebalance fault sample generator G for generating small deviation and a discriminator D for identifying true and false fault samples, initializing a weight value and a threshold value, inputting noise and a label into the generator G to generate low-proportion data set deviation-free fault samples, and inputting the generated fault samples and the true fault samples into the discriminator D for judgment until the discriminator D cannot judge whether the samples are true samples or generated samples;

step 3, establishing a cross-domain fault diagnosis classifier C, initializing a weight and a threshold, and simultaneously inputting a training set rebalancing fault sample and a test set fault sample to the fault classifier C;

step 4, the generator G simultaneously reduces L_gError and fault classifier error L_CThe discriminator D reduces L_DError, fault classifier C simultaneously reduces cross entropy loss L_clsAnd CORAL loss L_CORALReducing the deviation between the generated fault sample and the target domain sample data set;

and 5, carrying out fault diagnosis on the faults of the test set through the trained fault classifier C.

Further, in the step 4,

L_G＝L_g+L_C

L_C＝αL_cls+βL_CORAL

where α and β are constants between 0 and 1, i is the number of samples representing the ith sample, K is the number of training set samples, x_iIs the ith real sample, z_iIs the ith noise sample, c_iLabel of ith sample.

The method has high fault diagnosis accuracy, and is very suitable for fault diagnosis of the rolling bearing with unbalanced fault data and deviation of a data set.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.

FIG. 1 is a diagram of a rolling bearing fault diagnosis model for data imbalance and data set bias in an embodiment of the present invention;

FIG. 2 is a schematic diagram of a generator G according to an embodiment of the present invention;

FIG. 3 is a schematic structural diagram of a discriminator D according to an embodiment of the present invention;

FIG. 4 is a schematic structural diagram of a fault classifier C according to an embodiment of the present invention;

FIG. 5 is a flowchart of a rolling bearing fault diagnosis for data imbalance and data set bias in an embodiment of the present invention;

FIG. 6 is a graph of the results of fault diagnosis;

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

As shown in fig. 5, the embodiment of the present invention discloses a rolling bearing fault diagnosis method with unbalanced data and biased data set, which includes a generator G as shown in fig. 2, a discriminator D as shown in fig. 3 and a cross-domain fault classifier C as shown in fig. 4, where the fault diagnosis model is an end-to-end fault diagnosis model as shown in fig. 1. The method comprises the following steps:

step 1, collecting vibration signals of a rolling bearing as fault data, dividing the fault data into a training set and a testing set, wherein the training set fault sample and the testing set fault sample have deviation, the training set data contain labels, and the training set data are unbalanced; the test set data is balanced and does not contain labels;

specifically, the original failure data is normalized to the [0,255] interval, and the original signal is converted into a gray image.

Step 2, establishing a fault diagnosis model, wherein the fault diagnosis model comprises a rebalance fault sample generator G for generating small deviation and a discriminator D for identifying true and false fault samples, initializing a weight value and a threshold value, inputting noise and a label into the generator G to generate low-proportion data set deviation-free fault samples, inputting the generated fault samples and the real fault samples into the discriminator D for judgment, and continuing training the generator G until the discriminator D cannot judge whether the samples are the real samples or the generated samples if the generated fault samples and the real fault samples do not meet a termination condition. Generator G has a loss function of L_GThe discriminant D has a loss function of L_D. The generator G is used for generating fault samples with the smallest deviation as possible from the test set, and the generated fault samples are distributed the same as the real samples of the test set as possible; and the discriminator D is used for discriminating whether the sample is from the generated sample or the real sample, and rebalancing the original training fault sample.

And the generated samples and the real samples form a rebalancing training set fault sample, and the rebalancing training set fault sample and the test set fault data are transmitted to a fault classifier C. The generator G and the arbiter D are trained in a antagonistic manner for generating eligible rebalance training set fault samples.

Step 3, establishing a cross-domain fault diagnosis classifier C, initializing a weight and a threshold, simultaneously inputting a training set rebalancing fault sample and a test set fault sample to the fault classifier C, wherein the loss function of the fault classifier C is L_CThe generator loss function contains the fault classifier error L_CThe purpose is to reduce the data set bias of generating a fault sample from a target domain sample;

step 4, the generator G simultaneously reduces L_gError and fault scoreError L of analog device_CThe discriminator D reduces L_DError, fault classifier C simultaneously reduces cross entropy loss L_clsAnd CORAL (correlation alignment) loss L_CORALThe fault classifier C reduces the data set bias between the rebalance fault samples and the target domain samples, and the fault classifier error LC is fed back to the generator G for generating low-scale fault data with less data set bias. The fault diagnosis model adopts an end-to-end model to replace a traditional two-stage model, and directly generates a training set fault sample with small data set deviation;

and 5, carrying out fault diagnosis on the faults of the test set through the trained fault classifier C.

Further, in the step 4,

L_G＝L_g+L_C

L_C＝αL_cls+βL_CORAL

where α and β are constants between 0 and 1, i is the number of samples representing the ith sample, K is the number of training set samples, x_iIs the ith real sample, z_iIs the ith noise sample, c_iLabel of ith sample.

By adopting the invention, the CWRU (case Western Reserve university) rolling bearing database is subjected to fault diagnosis test, and the test result is shown in FIG. 6. The experimental result verifies the effectiveness of the method. The method has high fault diagnosis accuracy, and is very suitable for fault diagnosis of the rolling bearing with unbalanced fault data and deviation of a data set.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.

Claims (2)

1. A rolling bearing fault diagnosis method with unbalanced data and deviated data set is characterized by comprising the following steps:

step 1, collecting vibration signals of a rolling bearing as fault data, dividing the fault data into a training set and a testing set, wherein a fault sample of the training set and a fault sample of the testing set have deviation;

step 2, establishing a fault diagnosis model, wherein the fault diagnosis model comprises a rebalance fault sample generator G for generating small deviation and a discriminator D for identifying true and false fault samples, initializing a weight value and a threshold value, inputting noise and a label into the generator G to generate low-proportion data set deviation-free fault samples, and inputting the generated fault samples and the true fault samples into the discriminator D for judgment until the discriminator D cannot judge whether the samples are true samples or generated samples;

step 3, establishing a cross-domain fault diagnosis classifier C, initializing a weight and a threshold, and simultaneously inputting a training set rebalancing fault sample and a test set fault sample to the fault classifier C;

step 4, the generator G simultaneously reduces L_gError and fault classifier error L_CThe discriminator D reduces L_DError, fault classifier C simultaneously reduces cross entropy loss L_clsAnd CORAL loss L_CORALReducing the deviation between the generated fault sample and the target domain sample data set;

and 5, carrying out fault diagnosis on the faults of the test set through the trained fault classifier C.

2. The method for diagnosing the failure of the rolling bearing with the unbalanced data and the deviated data set according to claim 1, wherein in the step 4,

L_G＝L_g+L_C

L_C＝αL_cls+βL_CORAL

where α and β are constants between 0 and 1, i is the number of samples representing the ith sample, K is the number of training set samples, x_iIs the ith real sample, z_iIs the ith noise sample, c_iLabel of ith sample.

Images