TESIS: Member of the Vector Group
Virtual Test Driving

Your Contact

Neural networks for semi-active suspension control: analysis with virtual test drives

Use of neural networks in controller development with DYNA4

The DYNA4 Framework supports the simulation-based development of vehicle control systems, from MIL, SIL to HIL tests. For example, virtual test rigs for tires and axles provide automated workflows for the complex process of model parameterization. The example of a semi-active suspension control system based on artificial neural networks (ANN) illustrates how DYNA4 can be used to analyze control system behavior in a virtual test drive. The approach described here can be used equally well for many other control systems.

Overview of the control strategy

  • Optimization of ride comfort and driving safety based on model predictive control (MPC)
  • The explicit approach allows offline solving of time consuming optimal control problems
  • The use of neural networks enables nonlinear model-based control even in highly dynamic systems
  • Performance testing in the quarter car model and with full vehicle simulation using DYNA4 Car Professional
  • Real-world solution due to flexibly adaptable application parameters
  • Significant improvements in ride comfort and safety as a result

Model design and parameterization

  • Initial investigations of vertical dynamics and controller concepts in the quarter car
  • Parameters for the controller model derived from virtual component test rigs in DYNA4

Virtual Suspension Test Rig

- Usage of detailed MBS axles
- Detection of the (elasto-)kinematic axle behavior, e.g. the spring/damper ratio
Learn more >>

Load Distribution

- Investigation of the impacts of different load configurations
- Simple positioning using DYNAanimation
- Automated aggregation of mass inertia in DYNA4

Virtual Tire Test Rig

- Physically based FTire tire model, seamlessly integrated into DYNA4
- Tire parameterization using static and dynamic test cases on a virtual test rig
Learn more >>
 


Controller synthesis

  • Offline calculation of optimal control problems in the quarter car
  • Objective function: weighted combination of ride comfort and safety
  • Consideration of system limits
  • Damper current control based on vehicle state and vehicle characteristics

 

Controller evaluation in the full vehicle

  • Performance evaluation on realistic cobbled pavement road surface
  • Usage of the physically based FTire tire model
  • Test scenarios in DYNA4 on OpenDRIVE road with CRG surface
  • Significant improvement in ride comfort and safety using model predictive control (nonlinear programming - NLP) compared to reference control strategy (combined skyhook/groundhook - HH)

Find out more in the presentation slides of the chassis.tech 2017 conference or contact us directly.