Rotating machine fault information processing method, processing system, processing terminal and medium

Abstract

The invention belongs to the technical field of rotating machine state monitoring and fault diagnosis, and discloses a rotating machine fault information processing method, processing system, processing terminal and a medium, and the method comprises the steps: constructing a neural network model comprising a depth feature extractor, a domain classifier and a state predictor; automatically extracting migration fault features from laboratory simulation data and rotating part monitoring data in actual engineering equipment by using a depth feature extractor through a neural network model; utilizing a domain classifier to shorten the difference between two kinds of data distribution, utilizing a state predictor, introducing domain adaptation constraints, forming a fault diagnosis model based on a depth domain adaptive adversarial network, and utilizing the model to realize intelligent fault diagnosis of the rotating machinery. According to the method, migration fault features in laboratory simulation data and actual engineering data can be accurately extracted, a fault migration diagnosis model capable of being applied to the rotating part is formed, and an ideal effect is achieved through actual case utilization.

Description

technical field [0001] The invention belongs to the technical field of state monitoring and fault diagnosis of rotating machinery, and in particular relates to a method for processing fault information of rotating machinery, a processing system, a processing terminal, and a medium, and in particular to a method for processing fault information of rotating machinery based on a deep domain self-adaptive confrontation network . Background technique [0002] At present, rotating machinery, as one of the important components of large-scale industrial equipment systems, plays a vital role in the stable operation of the entire system. Its rotating parts (such as rolling bearings, gearboxes, etc.) often work in harsh environments, and have large rotational kinetic energy during operation, which is prone to failure, reduces system reliability, reduces system life, and even causes industrial production. huge loss. Therefore, it is very necessary to carry out fault diagnosis on rotat...

AI Technical Summary

Problems solved by technology

However, in practical engineering applications, two problems have been encountered: one is that the machine operates in a normal state for a long time, and the failure time only accounts for a very small part of the entire life cycle, which leads to an extreme imbalance between healthy samples and fault samples; The monitoring data during the long-term operation of machinery often does not contain fault information labels, and manual labeling is usually expensive, which leads to a serious shortage of labeled fault data samples

[0005] (1) Most of the traditional rotating machinery fault diagnosis methods are based on the time-frequency domain analysis of the fault mechanism and the experience and professional knowledge of diagnostic experts. The operation is cumbersome, and it is difficult to meet the current rapid analysis and diagnosis requirements for massive monitoring data.

[0006] (2) In the existing intelligent diagnosis technology for rotating machinery, due to the long-term operation of the machine in a normal state, the failure time only accounts for a very small part of the entire life cycle, resulting in an extreme imbalance between healthy samples and fault samples

[0007] (3) The monitoring data of machinery in the long-term operation process often does not contain fault information labels, while manual labeling is usually expensive, which leads to a serious shortage of labeled fault data samples, which is objectively data-driven intelligent fault diagnosis The application of the method to practical engineering presents challenges

[0008] The difficulty of solving the above problems and defects is as follows: In actual engineering case applications, due to the lack of fault samples and the difficulty of obtaining labeled data, the existing data-driven supervised fault diagnosis methods are difficult to solve during the model training process. It is prone to overfitting phenomenon due to lack of fault data, which leads to a decrease in the diagnostic accuracy of the model in the actual application process, and the generalization performance is seriously insufficient

Images