Generation of Rate-of-Injection (ROI) profile for Computational Fluid Dynamics (CFD) model of Internal Combustion Engine (ICE) using machine learning

•Rate of Injection (ROI) measurement and Post-processing.•Empirical model to predict ROI profile.•Machine learning (ML) models Random Forest (RF) and Neural Network (NN) to generate ROI profile.•CFD application using high-fidelity ROI profile generated by ML model. Injector configuration and spray characteristics are important parameters that define diesel combustion and emissions performance. One of the critical spray inputs is the Rate-of-Injection (ROI) profile. The ROI profile depends on the spray's operating conditions, including nozzle geometry (e.g., nozzle diameter), injection pressure, and injection duration. Besides, the internal nozzle flow phenomenon and external ambient conditions can further impact fuel introduction characteristics. This study measured the ROI profile of a heavy-duty (multi-hole) diesel injector using the Bosch tube technique. Injection pressure and injection duration were varied from 600 to 2600 bar and 0.5–3.0 ms, respectively. After post-processing, measurement data were then used to train numerical models, including a developed machine learning (ML) model that can create very similar ROI profiles with experimental data. Next, a Computational Fluid Dynamics (CFD) simulation used the ROI profile generated by ML model. For comparison, there are other simplified ROI profiles used in similar CFD simulation configuration. Results showed that the any difference in ROI profiles could affect the combustion and emissions significantly. This further emphasizes the need to provide high-fidelity spray input in terms of ROI profile for CFD simulation. The current ML model can deliver a realistic ROI profile for any given rail pressure and injection duration. [Display omitted]