Experimental Validation of an Innovative Approach for GDI Spray Pattern Recognition

In the present automotive scenario, along with hybridization, GDI technology is progressively spreading in order to improve the powertrain thermal efficiency. In order to properly match the fuel spray development with the combustion chamber design, using robust and accurate diagnostics is required. In particular, for the evaluation of the injection quality in terms of spray shape, vision tests are crucial for GDI injection systems. By vision tests, parameters such as spray tip penetration and cone angles can be measured, as the operating conditions in terms of mainly injection pressure, injection strategy, and chamber counter-pressure are varied. Provided that a complete experimental spray characterization requires the acquisition of several thousand spray images, an automated methodology for analyzing spray images objectively and automatically is mandatory. A decisive step in a spray image analysis procedure is binarization, i.e., the extraction of the spray structure from the background. Binarization is particularly challenging for GDI sprays, given their lower compactness with respect to diesel sprays. In the present paper, two of the most diffused automated binarization algorithms, namely the Otsu and Yen methods, are comparatively validated with an innovative approach derived from the Triangle method—the Last Minimum Criterion—for the analysis of high-pressure GDI sprays. GDI spray images acquired with three injection pressure levels (up to 600 bar) and two different optical setups (backlight and front illumination) were used to validate the considered algorithms in challenging conditions, obtaining encouraging results in terms of accuracy and robustness for the proposed approach.