Investigations on Multiple Injection Strategies in a Common Rail Diesel Engine Using Machine Learning and Image-Processing Techniques

The present study examines the effect of the multiple injection strategies in a common rail diesel engine using machine learning, image processing, and object detection techniques. The study demonstrates a novel approach of utilizing image-processing tools to gain information from heat release rates and in-cylinder visualizations from experimental or computational studies. The 3D CFD combustion and emission predictions of a commercial code ANSYS FORTE© are validated with small-bore common rail diesel engine data with known injection strategies. The validated CFD tool is used as a virtual plant model to optimize the injection schedule for reducing oxides of nitrogen (NO_x) and soot emissions using an apparent heat release rate image-based machine learning tool. A methodology of the machine learning tool is quite helpful in predicting the NO–soot trade-off. This methodology shows a significant reduction in soot and NO emissions using a pilot–main–post-injection schedule of 25% pilot, 25% post-, and 50% main injection, compared to a baseline pilot–main injection schedule. In addition, this work attempts a robust and high-fidelity optimization of the fuel injection schedule using the random forest algorithm for predicting the NO and soot emissions using 73 simulations done with different pilot–main and pilot–main–post-injection strategies on a small-bore diesel engine. Further, the object detection algorithm is trained on simulation data from the small-bore engine for detecting the interaction between the developed combustion from the pilot or main with sprays of subsequent injections using in-cylinder 3D CFD simulation and experimental data. A small-bore engine dataset shows that the trained object detection algorithm successfully corroborates the simulation and experimental data interaction. This investigation, therefore, presents a novel application of object detection methodology by automating the process and providing a general-purpose object detection algorithm. This approach can be used on any new simulation or experimental data for automated detection of the spray–thermal zone interaction without human intervention.