Research on Performance Degradation Prediction Method of Electric Vehicle Reducers Based on MDS-GA-SVR

Research on Performance Degradation Prediction Method of Electric Vehicle Reducers Based on MDS-GA-SVR

A method for modeling performance degradation with multiple dimensional scale (MDS) transformation and genetic algorithm optimized support vector regression (GA-SVR) is proposed to improve the prediction accuracy of electric vehicle reducers by fully exploiting the performance degradation informati...

Full description

Saved in:
Bibliographic Details
Main Authors: He Yinda, Li Weilin, Chen Feng, He Qingchuan, Pan Jun, Ma Xingjian
Format: Article
Language:zho
Published: Editorial Office of Journal of Mechanical Transmission 2024-01-01
Series:Jixie chuandong
Subjects:
Electric vehicle reducer
Multiple dimensional scaling
Genetic algorithm
Support vector regression
Degradation trend prediction
Online Access:http://www.jxcd.net.cn/thesisDetails#10.16578/j.issn.1004.2539.2024.01.020
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:A method for modeling performance degradation with multiple dimensional scale (MDS) transformation and genetic algorithm optimized support vector regression (GA-SVR) is proposed to improve the prediction accuracy of electric vehicle reducers by fully exploiting the performance degradation information. The features of vibration signals are extracted by using time domain, frequency domain, and time-frequency domain signal analyzing methods, and then the comprehensive degradation feature indicators are established by using the MDS algorithm. All the above-mentioned indicators are used as the data set for training and prediction. The optimal penalty parameter <italic>C</italic> and kernel parameter <italic>g </italic>are determined by using the genetic algorithm. A performance degradation model with high-precision is established based on the GA-SVR model by analyzing the testing data. The experiment results show that the prediction accuracy by using the proposed method is much higher than the results using PSO-SVR, GS-SVR and back propagation (BP) neural network. The RMSE values are reduced by 50.63%, 75.16% and 84.73%, and the <italic>R</italic><sup>2</sup> values are increased respectively by 3.93%, 6.51% and 9.51%, which proves the superiority of the proposed method.
ISSN:1004-2539

Similar Items