Rotary machine weak fault signal extraction method based on order analysis and sparse coding

Abstract

The invention discloses a rotary machine weak fault signal extraction method based on order analysis and sparse coding. The rotary machine weak fault signal extraction method comprises the following steps: firstly, transferring a constant rotating speed fault signal to an angle domain through order analysis; uniformly segmenting the angle domain signal, and training an angle domain displacement invariant dictionary by using displacement invariant sparse coding; further sparsely decomposing the variable speed angle domain fault signal containing strong noise by using a displacement invariant dictionary, and selecting a component with obvious impact characteristics according to a kurtosis criterion to reconstruct; and finally, extracting fault features through envelope spectrum analysis. Themethod can effectively eliminate the interference of the rotating speed and strong background noise, extracts the weak fault features under the variable rotating speed, and is high in flexibility andadaptability.

Description

technical field [0001] The invention belongs to the technical field of intelligent fault diagnosis of mechanical vibration signals, and relates to a method for extracting weak fault signals of rotating machinery based on order analysis and sparse coding. Background technique [0002] When the working surface of the parts in the rotating machinery is damaged, a periodic impact pulse signal will be generated during the operation, that is, the fault characteristic signal. Collecting the vibration signal during the operation of the rotating machinery through the vibration acceleration sensor, and performing related signal processing on the collected vibration signal can diagnose the corresponding faults of the rotating parts. At present, the research on fault feature extraction is mainly aimed at the operating state of stable speed conditions. However, in the actual operating environment, rotating parts often operate under variable speed conditions. Therefore, it is of practica...

AI Technical Summary

Problems solved by technology

For example, the Chinese patent application "A Regularized Sparse Filtering Method Applicable to Gear Fault Diagnosis at Variable Speed" (application number CN202010044852.6) uses the regularized sparse filtering method to automatically extract fault features and classify them through artificial neural networks. It needs to be pointed out Unfortunately, the neural network is highly dependent on training samples, and when the working conditions of the training samples are inconsistent with the working conditions of the test samples, it is easy to produce wrong diagnostic results.

For example, Beijing Tianze Zhiyun Technology Co., Ltd. applied for a patent "a method for extracting fault features of rotating machinery without a tachometer under variable speed" (application number CN201911285855.2), using sine or cosine waves to construct an over-complete dictionary to sparse the signal Decomposition, not self-adaptive, prone to under-decomposition, and the method of filtering by finding the peak of the spectrum is not suitable for the case where the amplitude of the fault signal itself is small

Images