An eddy current hardness sorting method based on ls-svm

Abstract

The invention discloses an eddy current hardness sorting method based on LS-SVM, which replaces the inverse operation of a large matrix in a conventional SVM algorithm by a finite number of iterative operations of a Lagrangian operator, and the solution process is a purely linear process, and more Easy to implement in hardware. The pure hardware algorithm has more advantages than the software algorithm in terms of speed due to the characteristics of processing parallelism, which provides the basis for the online sorting technology with high speed requirements on the production line; the acquisition of the training sample support vector coefficients The key is to determine the decisive parameters of the sorting hyperplane. The sorting method based on LS‑SVM can sort out qualified parts surrounded by unqualified parts. Its algorithm has strong practical value in the case that its eigenvalues ​​are quite different from those of other qualified products under the same furnace number, resulting in wrong sorting.

Description

technical field [0001] The invention relates to the technical field of hardness sorting, and more specifically relates to an LS-SVM-based eddy current hardness sorting method. Background technique [0002] Usually, the hardness sorting technology based on eddy current estimates the hardness value of the inspected object by extracting the impedance value of the eddy current detection coil. In order to correctly sort the inspected object, it is necessary to process the inspected object group in the same batch , the same model size, the same furnace number treatment and other conditions, after meeting this condition, use the hardness tester to find out the test pieces that meet the upper and lower boundary conditions in the tested object group, and use a series of software and hardware technologies to extract the detection coil Impedance value, according to the impedance value of the upper and lower boundary conditions, the sorting interval is drawn. If the impedance value of t...

AI Technical Summary

Problems solved by technology

For this method, there are the following problems in practical application: (1) There is a nonlinear relationship between the impedance value and the hardness value. If the hardness span of the object to be tested is large, the impedance value can represent multiple hardness values; (2) For the inspected objects processed in different furnace numbers and different batches, the upper and lower boundary conditions change in real time, and the mapping relationship between impedance value and hardness sorting interval is not easy to determine

Images