Neural network modelling of suspension dampers for variable temperature operation

A neural network model for predicting twin-tube hydraulic damper force is adapted for temperature varying operation. An efficient NARX model is presented with just two input variables: piston-rod velocity as the only input kinematic information, and second, the damper casing temperature. Although no internal heat generating mechanism is included in the model, an alternative simplifying assumption avoids the need for at least two additional input variables. The NARX network is trained using both measured and simulated random data and then appropriately tested for its generalisation capability. The paper shows that a simple NARX model, which uses damper casing temperature as an additional input variable, seems to characterise variable temperature operation very well. This model could be useful for both virtual testing, and for control studies involving adaptive suspension systems with passive dampers.