Energy Conversion and Management, Vol.177, 817-832, 2018
Methodology to simulate normalized testing cycles for engines and vehicles via design of experiments with low number of runs
This paper proposes a methodology for simulating engine/vehicle responses of a non-stationary test cycle by means of few steady-state operating modes, which can greatly reduce testing costs and time. The novelty of the proposed methodology is the application of a mapping from the engine working region to a square domain, which allows testing any design of experiments (DoE) regardless of the testing cycle, vehicle or engine. In this new working space (mapped region) it is easy to apply a DoE that satisfies optimality conditions. The validation of the methodology is based on experimental data obtained from the New European Driving Cycle. Firstly, the methodology consists in determining which representative responses can be instantaneously and/or cumulatively approximated via a low degree polynomial function (smooth surfaces) and, secondly, in performing a DoE analysis in the mapped working region where the points defining each DoE are placed. An approach function for each response is developed based on DoE tested points. Subsequently, this model allows simulating vehicle responses during the transient test. For the studied validation case, results show that main engine performance responses can be instantaneously and cumulatively predicted with high accuracy by means of a DoE with few runs. On the other hand, this analysis reveals a cumulative predicted response of confidence for regulated exhaust emissions, but not for the instantaneous values. These findings support future studies to determine the optimal DoE which minimizes testing time and costs with a satisfactory and accurate estimation of engine responses.