2024 Stuttgart International Symposium
on Automotive and Engine Technology

The calibration of Engine Control Units (ECUs) for road vehicles is challenged by stringent legal and environmental regulations, coupled with reduced development times. The growing number of vehicle variants, sharing similar engines and control algorithms, requires different calibrations. Additionally, these engines feature an increasing number of calibration parameters, along with complex parallel and nested conditions within the software, demanding a significant amount of measurement data during development. The current state-of-the-art logic-level (white box) model-based calibration proves effective but involves considerable effort for model construction and validation. This is often hindered by limited functional documentation, available measurements, and hardware representation capabilities. This article introduces a model-based calibration approach using Neural Networks (black box) for two distinct ECU functional structures with minimal software documentation. The ECU is operated on a Hardware-in-the-Loop (HiL) rig for measurement data generation, supplemented with real-world test drive measurements. To build surrogate models of ECU functions, Neural Network model inputs are categorized into two: function inputs as perceived by the logic level (white box) software function, and curve/map fitting parameters representing the calibratable parameters of the ECU function. Factors influencing surrogate model accuracy, such as Neural Network hyperparameter optimization, input space amount and distribution as well as the optimization algorithms are investigated. Results show scalability of ECU function model representation with measurement data and increased accuracy with the number of implemented parameters. The proposed approaches demonstrate semi-automatic calibration with high accuracy and robustness. Robustness analysis indicates insensitivity to initial values in single map optimization but variability in optimization results for multiple maps due to inter-map correlations. In addition to calibration, the presented function representation method facilitates the use of plant models to replace time-consuming function construction and validation.