A fault diagnosis method based on deep convolutional domain adversarial transfer learning

Abstract

The invention discloses a fault diagnosis method based on deep convolution domain confrontation transfer learning. In DCDATL, a deep convolution residual feature extractor is used for high-level feature extraction, which improves the convergence and nonlinear approximation ability of DCDATL; The Kronecker product of high-level features and label information obtains the feature joint distribution representation and embeds it in the domain classifier, and conducts domain adversarial training to improve the transfer performance of DCDATL; feature transfer based on minimizing DCDATL's joint distribution domain adversarial overall loss function And the classification process improves the classification accuracy after transfer. The above advantages of DCDATL enable the fault diagnosis method based on DCDATL to use the labeled samples under the historical working conditions to perform high-precision fault diagnosis on the rotating machinery currently under test when the labeled samples under the current working conditions of the rotating machinery do not exist.

Description

technical field [0001] The invention belongs to the technical field of fault diagnosis methods for rotating machinery, and in particular relates to a fault diagnosis method based on deep convolution domain confrontation transfer learning. Background technique [0002] Rotating machinery equipment is widely used in metallurgy, aviation, transportation, chemical and energy industries, and its mechanical structure is developing in the direction of large-scale, heavy-duty, precision and high-speed. Condition monitoring and fault diagnosis are important means to ensure long-term safe and reliable operation of modern industrial equipment. [0003] Due to the fact that rotating machinery often works in an industrial environment with variable working conditions during the entire service process, the probability of failure is high and it is easily damaged. In the environment of variable working conditions (such as different rotational speeds, different loads), there is a problem tha...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the above-mentioned deficiencies in the prior art, the fault diagnosis method based on deep convolutional domain adversarial transfer learning provided by the present invention solves the problem of obtaining a large number of labeled samples of rotating machinery in the existing rotating machinery fault diagnosis method (that is, known Fault type samples) are relatively difficult, so that there are fewer labeled samples under the current working conditions, resulting in a problem of low fault diagnosis accuracy for the current samples to be tested

Images