SUBARU Uses AI Surrogate Model to Reduce Transmission Control System Analysis Time

The AI model can now reproduce waveforms at any source pressure, oil temperature, and current. The calculation time can be significantly reduced while ensuring the accuracy of hydraulic waveforms.

Key Outcomes

  • Achieved a 99% reduction in calculation time compared to the original 1D model
  • Constructed AI surrogate model in MATLAB that can reproduce waveforms with arbitrary current, oil temperature, and source pressure readings
  • Accurately reproduced waveforms, even in oil temperature ranges where the model has not been trained
Video length is 19:45

In automatic transmission vehicles, the transmission hydraulic control system regulates the flow of hydraulic fluid, ensuring smooth gear shifts and efficient power delivery in different driving conditions. Analyzing and optimizing the hydraulic control system can involve using large-scale, detailed physical simulation models, which can be computationally intensive and time-consuming. At SUBARU, engineers are addressing this challenge by employing an AI surrogate model.

The AI surrogate was constructed with MATLAB® using a neural ODE model. By applying this AI surrogate model, SUBARU achieved a significant reduction in calculation time compared to a third-party 1D physical model analysis.

While providing reduced calculation time, the AI model maintained the waveform accuracy of hydraulic pressure, even in oil temperature ranges where it has not been trained.