Adaptive DOE-Based Optimization tool for high performance Engine Calibration

This thesis presents the development of a methodology and a dedicated tool for the calibration and optimization of high-performance spark-ignition engines through the combined use of experimental testing and Design of Experiments (DoE) techniques. The work begins with an overview of the engine testing environment, describing the operation of dynamometers, the measurement chain and the subsystems required to reproduce controlled boundary conditions with high repeatability. Building on this foundation, the classical calibration workflow is analyzed in detail, from preliminary hardware verification to the characterization of motored and fired operation, base engine mapping and the definition of performance, emission and technological constraints that govern the optimization problem. A critical assessment of traditional One-Factor-At-a-Time (OFAT) calibration highlights its intrinsic limitations in handling multivariable interactions, motivating the transition toward more advanced model-based strategies. The thesis therefore introduces the theoretical principles of DoE, discussing domain definition, design selection, design quality metrics and local modeling approaches including polynomial regression, radial basis functions, Gaussian processes and artificial neural networks. To validate these concepts, a synthetic optimization environment is implemented to benchmark different algorithms and assess their ability to identify optimal parameter combinations under constrained multidimensional spaces. The second part of the work focuses on the design and implementation of an experimental optimization tool aimed at improving calibration efficiency on the test bench. The tool integrates operating-point definition, iterative DoE construction, geometric and stochastic sampling strategies, surrogate-model generation and automated optimization routines. Particular attention is given to domain expansion, model refinement and the management of actuator sensitivities. The resulting workflow enables the calibration engineer to combine structured experimental planning with domain knowledge in order to reduce test time, improve model accuracy and obtain a globally consistent engine map. Overall, this thesis demonstrates how a systematic, model-based approach can significantly enhance the efficiency, robustness and transparency of the calibration process, providing a practical framework for modern powertrain development where reduced timelines, increasing system complexity and stringent regulatory constraints demand more advanced calibration methodologies.