Gearbox state fault diagnosis method based on multi-scale multi-source heterogeneous information fusion

A multi-source heterogeneous and fault diagnosis technology, applied in neural learning methods, machine/structural component testing, neural architecture, etc., can solve problems such as vibration signal acquisition system failure, large noise, and CNN feature extraction capabilities need to be improved

Pending Publication Date: 2021-12-31
NORTHWESTERN POLYTECHNICAL UNIV
View PDF0 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0002] The currently widely used fault detection system is based on the vibration signal for diagnosis, and the vibration signal is collected by the acceleration sensor installed in the shell. The reliability of the fault d

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Gearbox state fault diagnosis method based on multi-scale multi-source heterogeneous information fusion
  • Gearbox state fault diagnosis method based on multi-scale multi-source heterogeneous information fusion
  • Gearbox state fault diagnosis method based on multi-scale multi-source heterogeneous information fusion

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0098] Data Layer Fusion

[0099] By reading the infrared image and zooming to get its uniform size digital matrix, its form is [64,32,3], and the vibration information with a length of 2048 is normalized into a grayscale image, the specific form is [64,32,1 ], the two matrices are fused in the last dimension, that is, the channel, and reshaped into a matrix of [64,32,4] as the input of the neural network.

[0100] feature extraction

[0101] Input a fused image of 64×32×4 on the first scale, go through 16 convolution kernels with a kernel size of 3×3, select a step size of 1×1, and obtain a convolutional layer output of 64×32×16;

[0102] The Relu activation function is used to activate and output; after the maximum pooling layer with a kernel size of 2×2 and a step size of 2×1, a pooling layer output of 32×32×16 is obtained;

[0103] After repeating the above-mentioned convolution, activation, and pooling twice, a 4×4×4 feature is obtained;

[0104] On the second scale, i...

Embodiment 2

[0111] Data Fusion

[0112] Convert the read infrared image into a grayscale image, the size is [64,32,1], normalize the vibration information with a length of 2048 and convert it into a grayscale image, the specific form is [64,32,1] , to realize the fusion of the two matrices, and reshape the matrix of [64,32,2] as the input of the neural network.

[0113] feature extraction

[0114] Input a fused image of 64×32×2 on the first scale, pass through 16 convolution kernels with a kernel size of 3×3, select a step size of 1×1, and obtain a convolutional layer output of 64×32×16;

[0115] The Relu activation function is used to activate and output; after the maximum pooling layer with a kernel size of 2×2 and a step size of 2×1, a pooling layer output of 32×32×16 is obtained;

[0116] After repeating the above-mentioned convolution, activation, and pooling twice, a 4×4×4 feature is obtained;

[0117] On the second scale, input a fusion image of 32×16×4, and obtain the features of...

Embodiment 3

[0124] Data Fusion

[0125] Convert both the infrared image and the vibration signal into a grayscale image with a size of [64,32,1], and then perform horizontal stitching, and the horizontal stitching is a matrix of [64,64,1], that is, to realize the second dimension, namely Fusion over channels, reshape to a matrix of [64, 32, 2] as input to the neural network.

[0126] feature extraction

[0127] Input a fused image of 64×64×1 on the first scale, pass through 16 convolution kernels with a kernel size of 3×3, select a step size of 1×1, and obtain a convolutional layer output of 64×32×16;

[0128] Use the Relu activation function to activate and output; after the maximum pooling layer with a kernel size of 2×2 and a step size of 2×2, a pooling layer output of 32×32×16 is obtained;

[0129] After repeating the above-mentioned convolution, activation, and pooling twice, a 4×4×4 feature is obtained;

[0130] On the second scale, input a fusion image of 32×16×4, and obtain the...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses a gearbox state fault diagnosis method based on multi-scale multi-source heterogeneous information fusion, which comprises the following steps: acquiring and preprocessing an infrared image and a vibration signal when a gearbox has a fault, performing different-scale division on the preprocessed infrared image and the vibration signal by adopting an improved multi-scale decomposition mode based on a Gaussian pyramid, converting infrared images and vibration signals of different scales into grey-scale maps of corresponding sizes, fusing image data, inputting the fused data into a convolutional neural network, extracting gearbox fault features under different scales, and constructing a multi-scale feature fusion fault diagnosis model, inputting an infrared image and a vibration signal of the gearbox in the multi-scale feature fusion fault diagnosis model, and outputting a fault type of the gearbox. According to the method, data features of original image transformation can be extracted in parallel, and multiple fault modes of the gearbox can be identified through an end-to-end learning method.

Description

technical field [0001] The present invention relates to the technical field of automatic diagnosis of mechanical faults, and more specifically relates to a gearbox state fault diagnosis method based on fusion of multi-scale, multi-source and heterogeneous information. Background technique [0002] The currently widely used fault detection system is based on the vibration signal for diagnosis, and the vibration signal is collected by the acceleration sensor installed in the shell. The reliability of the vibration signal sensor for fault diagnosis is low. On the other hand, the vibration signal acquisition system may also fail. In addition, under complex working conditions, the feature extraction ability of CNN needs to be improved. This is because the simple CNN structure is difficult to extract effective features because the object data is too complex, so it is necessary to optimize the CNN structure to improve the feature extraction ability of CNN. Contents of the invent...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
IPC IPC(8): G01M13/021G01M13/028G06K9/62G06N3/04G06N3/08
CPCG01M13/021G01M13/028G06N3/08G06N3/045G06F18/2415G06F18/253
Inventor 李永波杨玉龙乔彬王欣悦李霓布树辉邓子辰张凯贾思详孙丁一
Owner NORTHWESTERN POLYTECHNICAL UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap