A Rotating Machinery Health Assessment Method Based on Deep Autoencoder Networks

Abstract

The invention discloses a rotating machinery health assessment method based on a deep self-encoding network, including step 1, vibration signal collection; step 2, original feature extraction; step 3, using a deep self-encoding network DAE to perform feature dimensionality reduction; step 4, feature Selection; step 5, use unsupervised SOM algorithm to construct health indicators; step 6, use genetic algorithm-based fusion evaluation criteria to evaluate health indicators. The present invention combines the advantages of the powerful feature extraction ability of deep learning, and combines the deep self-encoding and the minimum quantization error method. In addition, aiming at the problem that the evaluation criterion based on one metric often has deviation, a fusion evaluation criterion based on genetic algorithm is provided. The invention can accurately evaluate the health status of rotating machinery, can be widely used in the health assessment of rotating machinery in the fields of chemical industry, metallurgy, electric power, aviation, etc., can accurately describe the dynamic process of performance degradation of these components, and can also perform remaining life prediction.

Description

technical field [0001] The invention relates to a rotating machinery health assessment technology, in particular to a rotating machinery health assessment method based on a deep self-encoding network. Background technique [0002] Due to the development of advanced sensor and computer technology, a large amount of condition monitoring data has been accumulated in industrial production, and data-driven methods have been widely used in bearing prediction because they are able to use condition monitoring data to quantify the degradation process instead of establishing a Accurate system models are not readily available. [0003] Generally, a data-driven prediction method usually consists of the following three steps: data acquisition, health indicator construction, and remaining life prediction. Health indicators attempt to identify and quantify historical and ongoing degradation processes by extracting characteristic information from acquired data. Therefore, the quality of t...

AI Technical Summary

Problems solved by technology

However, PCA is a linear dimension reduction method, and the bearing degradation process is a nonlinear degradation process, so it cannot be used to accurately evaluate the health indicators of rotating machinery

In addition, the health indicators constructed by the combination of PCA and time-frequency domain feature extraction methods need to be improved in terms of monotonicity, trend, and robustness

Images