A quantitative method of calculating transient nonlinear heat partition coefficient between clutch friction discs with deep learning

Peng Zhang, Changsong Zheng, Cenbo Xiong*, Biao Ma, Liang Yu, Dengming Luo

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

Abstract

In the current clutch temperature field study, the generally used constant heat partition coefficient tends to overestimate the separator disc temperature and underestimate the friction disc temperature. Although some researchers have found the characteristics of the time-varying heat partition coefficient, a suitable method is still needed to apply it to temperature calculations. This study provides a quantitative method for the application of the transient nonlinear heat partition coefficient to temperature calculations. The finite difference method is adopted to figure out the time-varying curve of the heat partition coefficient by coupling the contact temperature of the friction components. The numerical results show that the heat partition coefficient is independent of rotation speed with three stages: initial value, rapid time-varying, and steady-state. Different from the analytical method, we apply a deep learning method to train the quantisation function to characterise these three stages, avoiding complex formula derivation. As a result, the quantitative function can characterise the time-varying heat partition coefficient accurately, with an average error of 0.19%, 3.05% and 0.62% for the inert, time-varying, and steady-state stages, respectively. In addition, the accuracy of applying the quantisation function in temperature simulation is verified by friction experiments, and the error is less than 8%. This is superior to the results of solving the temperature field by a constant heat partition coefficient.

Keywords

  • Clutch
  • deep learning
  • heat partition coefficient
  • quantification function
  • temperature field

Fingerprint

Dive into the research topics of 'A quantitative method of calculating transient nonlinear heat partition coefficient between clutch friction discs with deep learning'. Together they form a unique fingerprint.

Cite this