|
CSME 2025/04
Volume 46 No.2
:
141-151
|
|
Observational Modeling of Clutch Thermal Effects and Temperature Protection Prediction with Long and Short-Term Memory Network
Xiao-Hu Geng a, Wei-Dong Liu b and Yu-Long Lei c
aGraduate Student, State Key Laboratory of Automotive Simulation and Control, Jilin University, Changchun, Jilin 130000, China
bProfessor, State Key Laboratory of Automotive Simulation and Control, Jilin University, Changchun, Jilin 130000, China.
cProfessor, State Key Laboratory of Automotive Simulation and Control, Jilin University, Changchun, Jilin 130000, China.
|
Abstract:
When an automatic mechanical transmission (AMT) shifts gears frequently, the clutch is repeatedly in the state of slipping, which makes the clutch thermal effects accumulate rapidly. The temperature in the frictional parts rises and spreads after the clutch's frictional work is applied, that needs some moments. To observe the clutch thermal effects and overcome the delay of temperature rise variation on clutch control, this paper proposes a prediction control for clutch temperature protection based on the long shortterm memory (LSTM) network. Firstly, an extended state observation (ESO) model for clutch temperature based on clutch dynamics and heat transfer theory was established to estimate the temperature of the friction parts. Secondly, the future trend of clutch temperature was predicted based on the LSTM network algorithm, and the temperature was prevented from exceeding the threshold by adjusting the shift frequency and clutch input torque. Finally, the bench experiment was conducted to manipulate the AMT and clutch for 20 upshift and downshift cycles to compare the effect of temperature protection with LSTM and recurrent neural network (RNN) predictive control. The results shown the 16.21% reduction in high temperature operating time (over 250°C) with LSTM predictive control and the friction work was reduced from 5812.47kJ to 5380.48kJ.
|
|
Keywords: clutch thermal effects, temperature prediction, long and short-term memory network
|
|
|
*Corresponding author; e-mail:
|
©
2025
CSME , ISSN 0257-9731
|