Method for extracting features of crack acoustic emission signal of drawing part

Abstract

The invention discloses a method for extracting features of a crack acoustic emission signal of a drawing part. The method comprises the following steps of: first, preprocessing an acquired original acoustic emission signal in a computer; then, performing empirical mode-based decomposition on the preprocessed acoustic emission signal to obtain n intrinsic mode function components and a residual component; next, performing Hilbert transform on each intrinsic mode function component and expressing the amplitude of the signal as a local wave time-frequency spectrum in Hilbert space; later on, dividing the plane of the local wave time-frequency spectrum into m regions equally, respectively calculating local energy of a time-frequency domain of each region, normalizing the local energy of the time-frequency domain of each region and taking the normalized local energy of the time-frequency domain as an initial feature parameter; and finally, performing a genetic algorithm operation on the initial feature parameter after a plurality of iterations, and obtaining an optimal feature parameter by realizing automatic reorganization and optimization on the initial feature parameter. By using the method, interferences caused by other components are eliminated; the signal-to-noise ratio is improved; the optimal feature parameter can be searched quickly; the diagnostic time can be shortened remarkably; and the diagnostic efficiency can be improved.

Description

technical field [0001] The invention relates to a feature extraction method of crack acoustic emission signals of deep-drawing parts, which can be applied to the quality monitoring of deep-drawing parts in the process of metal State identification and quality monitoring are carried out in the extrusion forming process. Background technique [0002] During the forming process, the deep-drawn parts not only have to bear high contact pressure and severe friction, but also the periodic changes of stress, strain and temperature caused by cyclic loading, which will cause cracks in the parts. Some tiny cracks are difficult to detect with the naked eye. During the production process There may be batches of waste products in the process, which has brought huge economic losses to the enterprise. At present, the quality of deep-drawing parts is judged entirely by on-site manual experience, which belongs to post-event inspection and has a certain time lag. In fact, when cracks are fou...

AI Technical Summary

Problems solved by technology

The disadvantage of this method is that the wavelet packet transform needs to select the wavelet base and the number of decomposition layers, which is not self-adaptive, and because the crack acoustic emission signal is a short-term impact signal, it cannot meet the stability requirements of time series modeling. Difficult to extract crack acoustic emission signal features is actually difficult to achieve

The disadvantage of this method is that the training of the neural network needs to collect a large amount of data to further optimize the structure of the BP network, otherwise it will affect the speed and accuracy of network recognition, and it is time-consuming and labor-intensive to apply on the production site, making it difficult to achieve fast and accurate diagnosis the goal of

Images