|
|
|
| Study on Regression Prediction Method of Multi signal Spindle PSO-SVM Rotation Error Based on LMD |
|
|
|
|
Abstract In order to solve the problem that the original signal type used in the prediction of spindle rotation error of CNC machine tools is too single and the eigenvalue selection is insufficient, this paper adopts a regression prediction method of spindle rotation error based on the combination of multiple signals and eigenvalue optimization algorithm. Firstly, the complex vibration signal, current signal and acoustic emission signal of the main shaft are decomposed into the sum of several product function (PF) components with physical significance by using local mean decomposition (LMD) method; Secondly, the time-frequency domain eigenvalues of the main PF components (product function components) in the three signals are extracted; The Pearson coefficient between the rotation error and the eigenvalue is calculated to optimize several eigen-values, and the eigenvalue with the strongest correlation with the spindle rotation error in each PF component is obtained, which is used as the input parameter to predict the spindle rotation error; Finally, the spindle rotation error is accurately predicted by establishing the support vector machine prediction model from the eigenvalue to the spindle rotation error. The results show that when the three spindle signals are used as inputs, the rotation error prediction model is better. At the same time, combined with the optimization algorithm of eigenvalue, the effect of the particle swarm optimization support vector machine rotation error prediction model is the best. The mean square error, determination coefficient, and average absolute error of the test set sample set can reach 0.27%, 0.9094, 4.48%. It provides an effective method for on-line pre-diction of spindle rotation error.
|
|
Received: 07 October 2023
Published: 26 September 2024
|
|
|
|
|
|
|
2024, Vol. 45 
