Sequential DoE framework for steady state model based calibration

Abstract

The complexity of powertrain calibration has increased significantly with the development and introduction of new technologies to improve fuel economy and performance while meeting increasingly stringent emissions legislation with given time and cost constraints. This paper presents research to improve the model-based engine calibration optimization using an integrated sequential Design of Experiments (DoE) strategy for engine mapping experiments. This DoE strategy is based on a coherent framework for a model building - model validation sequence underpinned by Optimal Latin Hypercube (OLH) space filling DoEs. The paper describes the algorithm development and implementation for generating the OLH space filling DoEs based on a Permutation Genetic Algorithm (PermGA), subsequently modified to support optimal infill strategies for the model building - model validation sequence and to deal with constrained non-orthogonal variables space.

The development, implementation and validation of the proposed strategy is discussed in conjunction with a case study of a GDI engine steady state mapping, focused on the development of an optimal calibration for CO₂ and particulate number (Pn) emissions. The proposed DoE framework applied to the GDI engine mapping task combines a screening space filling DoE with a flexible sequence of model building - model validation mapping DoEs, all based on optimal DoE test plan augmentation using space filling criteria. The case study results show that the sequential DoE strategy offers a flexible way of carrying out the engine mapping experiments, maximizing the information gained and ensuring that a satisfactory quality model is achieved.