Barkhausen signal stress estimation method based on slow feature analysis

Abstract

The invention discloses a Barkhausen signal stress estimation method based on slow feature analysis. Barkhausen signals under different stresses are used, Barkhausen slow signals are extracted througha slow feature analysis algorithm, and eigenvalues of the Barkhausen slow signals are extracted, and finally the obtained eigenvalues are used to fit stress data. Because residual stress is an inherent property of a material, the slow feature analysis method has higher accuracy. The method comprises the following steps: step 1, combining Barkhausen signals under different stresses into a vector;step 2, performing slow feature analysis on the combined Barkhausen signal to obtain the Barkhausen slow signal; step 3, extracting an upper envelope of the Barkhausen slow signal, and obtaining a Barkhausen signal slowly varying feature; step 4, averaging the slow feature of the Barkhausen slow signal, to form a feature vector; step 5, performing linear regression on different stresses and eigenvectors to obtain an expression model of features and stress of the Barkhausen slow signal.

Description

technical field [0001] The invention relates to a Barkhausen signal stress estimation method based on slow feature analysis, which is suitable for quantitative and non-destructive evaluation of the stress of ferromagnetic materials. Background technique [0002] Ferromagnetic materials, as engineering basic materials, will be subjected to stress in different directions during manufacture and service, resulting in residual stress, which will lead to a decline in the performance of ferromagnetic materials; Fracture will affect the safety of production work. In order to ensure the safe, long-term and stable operation of industrial production, it is necessary to conduct quantitative non-destructive evaluation of the stress conditions of ferromagnetic materials in service. [0003] For the detection of residual stress, traditional detection techniques mainly include X-ray diffraction, metallographic analysis and hardness measurement methods. These traditional technologies will ...

AI Technical Summary

Problems solved by technology

The traditional eigenvalues ​​include: the average value of the Barkhausen signal, the root mean square value, the peak value of the envelope, the number of pulses, etc., but when these eigenvalues ​​are used to fit the stress, they all have problems such as poor linearity and high dispersion

Images