New Hybrid Genetic Algorithm for Pitch Sequence Optimization of CVT Variator Chain

Paper #:
  • 2017-01-1120

Published:
  • 2017-03-28
Abstract:
A CVT variator chain system is superior in efficiency to a variator belt system because of its lower internal friction. However, a chain produces more noise than a belt due to the long contact length between the rocker pins and pulley sheaves. This paper focuses on optimization of the link arrangement pattern (pitch sequence) in order to reduce chain noise. The current pitch sequence puts links of different lengths together appropriately and reduces the peak level of 1st-order chain noise by improving noise dispersibility. First, the object function for optimization was defined as the reduction of the peak level of 1st-order chain noise combined with a well-balanced chain noise level of low and high frequency bands. Road noise and wind noise as transmission background noise have the characteristic that they increase as the frequency decreases. Therefore, the object function was aimed at reducing the peak level of 1st-order chain noise by moving chain noise energy to the low frequency side from the high frequency side in consideration of the masking effect of the background noise. Clustering was then used to visualize the design space of the pitch sequence in order to choose an optimization method suitable for the characteristic of the pitch sequence design space. This design space was clarified as being a multimodal space. Finally, a new optimization method called Hybrid-GA, which combines a genetic algorithm with local search, was developed to find the most suitable solution of this multimodal design effectively. Hybrid-GA found the most suitable pitch sequence from a very large design space of "3 to the 80th power" patterns. The chain noise of this sequence is lower than that of the reference sequence by 2.8 dB at the peak level of 1st-order chain noise.
Access
Now
SAE MOBILUS Subscriber? You may already have access.
Buy
Attention: This item is not yet published. Pre-Order to be notified, via email, when it becomes available.
Select
Price
List
Download
$22.00
Mail
$22.00
Members save up to 36% off list price.
Share
HTML for Linking to Page
Page URL

Related Items

Training / Education
2017-08-15
Article
2016-02-02
Article
2016-02-02
Article
2016-02-02