Automatic Calibration Strategy Based on Artificial Neural Networks for Shift Control of Automatic Transmission

As the number of gears in automatic transmissions (AT) increases, the calibration parameters in the gear shift control process of the transmission control unit (TCU) increase exponentially, significantly increasing the calibration workload during engineering development. To address the challenges of high cost and long cycle times associated with traditional manual calibration, this paper proposes an automatic calibration strategy for shift control based on artificial neural networks (ANNs). The core of this method lies in utilizing an ANN to establish a non-linear mapping relationship between shift characteristics and calibration parameters, thereby simulating and replacing the analysis and adjustment process of engineers. In this research, a vehicle simulation model based on a 9-speed automatic transmission (9AT) was first constructed. A large-scale dataset of shift characteristics was obtained by traversing various parameter combinations, and key features were extracted for model training. Simulation results demonstrate that the trained ANN model performs excellently in the automatic calibration process, requiring only 4 to 5 iterations to adjust shift quality to a level comparable to manual calibration. Its convergence speed and efficiency are significantly superior to traditional rule-based calibration methods. Furthermore, the model exhibits a certain degree of generalization ability and robustness across different throttle openings and gear-shifting conditions. The proposed automatic calibration method does not rely on high-precision physical models, effectively shortening the development cycle and improving calibration efficiency, which holds significant application value in the field of automatic transmission engineering development.