Core Concepts
The author proposes a data-based model combining Principle Component Decomposition and Gaussian Process Regression to predict in-cylinder pressure and cycle-to-cycle variations for combustion control.
Abstract
The study focuses on developing a model to predict in-cylinder pressure and cycle-to-cycle variations using a data-based approach. The model combines Principle Component Decomposition and Gaussian Process Regression, showing promising results for combustion control applications.
Over the years, various models have been proposed to predict combustion measures or full in-cylinder pressure, but they often fail to capture cycle-to-cycle variations. The proposed approach aims to address this limitation by modeling in-cylinder pressure based on in-cylinder conditions and fuel settings.
The study demonstrates the effectiveness of the model on an RCCI engine running on Diesel and E85 fuels, showcasing good accuracy in predicting combustion measures like Gross Indicated Mean Effective Pressure (IMEPg) and peak-pressure rise-rate.
By analyzing different hyperparameters and kernel choices, the study provides insights into optimizing the model's performance for advanced combustion concepts with large cycle-to-cycle variation.
The research contributes to advancing control-oriented combustion models that are crucial for improving efficiency, safety, and performance of internal combustion engines.
Stats
The prediction quality of the evaluated combustion measures has an overall accuracy of 13.5% and 65.5% in mean behavior and standard deviation, respectively.
The peak-pressure rise-rate accuracy is 22.7% in mean behavior and 96.4% in standard deviation.