Mechanical fault feature extraction method based on wavelet transform and topological data analysis

Abstract

The invention relates to a mechanical fault feature extraction method based on wavelet transform and topological data analysis, and the method comprises the steps: carrying out the feature analysis of an original signal through the wavelet transform and topological data analysis, carrying out the parallel fusion of two groups of features, and inputting the two groups of features into a subsequent classification network; on the one hand, wavelet transform analysis is adopted to consider the time domain and frequency domain characteristics of mechanical fault data, original vibration signals are analyzed from multiple scales, and better characteristics are extracted from non-stationary original time domain signals; on the other hand, topological data analysis is applied to the field of fault diagnosis, the deep topological relation between fault data sets and the shape in hidden data can be effectively mined out through the topological data analysis method under the influence of noise, more robust and more important data features are extracted, and the subsequent fault type classification accuracy can be improved.

Description

technical field [0001] The invention relates to a mechanical fault feature extraction method based on wavelet transform and topological data analysis, and belongs to the technical field of feature extraction. Background technique [0002] With the rapid development of science and technology, the functions of mechanical equipment in modern industry are becoming more and more complex. Fault diagnosis of mechanical equipment becomes the most critical aspect in system design and maintenance. Machine failures can cause huge financial losses and sometimes pose a threat to the people who use the machines. Fault diagnosis plays an important role in monitoring the relationship between data and machine health status, which has become a widespread concern in machine health management, where improving diagnostic accuracy is particularly important. In the fault classification step, feature extraction is the most critical part. Therefore, the quality of features extracted from mechanica...

AI Technical Summary

Problems solved by technology

[0004] However, on the one hand, the diagnostic performance of most traditional methods depends largely on the quality of the extracted and selected features. In engineering practice, the collected vibration signals are always data-heavy, complex and unstable, and noisy very big

Not only that, but in most cases, feature selection largely depends on the engineering experience of diagnostic experts. For complex and large data sets, it becomes very difficult to use expert experience to make useful assumptions.

On the other hand, it is difficult for shallow learning models to effectively learn complex nonlinear relationships among fault data.

Images