Time-domain-based model correlation and validation using state-parameter-disturbance estimation techniques.

Reducing emissions and pollution in all aspects of daily life is high on the EU’s agenda. Road transport is also complying with strict emissions regulations, paving the way with electric vehicles (EVs), internal combustion engines and hybrids eco-powertrains. However, there is an urgent need to test and simulate several complex challenges related to performance. This PhD focuses on time domain identification methods for modelling mechatronic systems. Standard methods focus on optimizing parameters by either minimizing the discrepancy between the time history of the measurement signal and the model output, or by optimizing towards a certain global criteria, i.e., shifting time, transmission error in an ad hoc fashion. This often results in a fit based on the training data, and does not necessarily offer good predicting capabilities for operating regimes outside the regimes of the training set. Consequently, there is a clear need for novel identification strategies that not only optimize the parameters but also iterate on the equation structure and finally indicate which equations are responsible for the remaining model mismatch. An additional challenge is linked to incorporating different measurement sets and different operational conditions in a single identification step. The focus is on powertrains where the distributed flexibility is dominant and several non-linear aspects are present, i.e., velocity-dependent stiffening effects, Coriolis forces due to bending loads, load-dependent bearing stiffness. The result will be a pre-test correlation and validation approach for non-linear systems.