Research on knock recognition of coal-based naphtha homogeneous charge compression ignition engine based on combined feature extraction and classification

Abstract

The purpose of this study is to investigate combustion characteristics and knock recognition of HCCI engines fueled with coal-based naphtha by analyzing in-cylinder pressure. The in-cylinder pressures of knock and normal combustion at two different intake temperatures (Tin) of 60 °C and 90 °C were collected on a modified diesel engine, characteristic vector was constructed from the in-cylinder pressure signal based on empirical mode decomposition (EMD) and sample entropy (SampEn), and a support vector machine (SVM) classifier was used to recognize knock. The results show that in-cylinder pressure, heat release rate (HRR), in-cylinder temperature, pressure rise rate (PRR) and pressure rise acceleration (PRA) change steadily in normal combustion. With occurrence of knock, they have a sharp rise and huge fluctuations. HRRmax, PRRmax, and PRAmax are greatly increased, and the corresponding phase is significantly advanced, especially in high-temperature stage. When Tin is 90 °C, PRAmax reaches 1.174 MPa/°CA2, which is an increase of 854.47% compared to normal combustion. However, the coefficient of variation for peak pressure (COVPmax) of two Tins show completely opposite results. At Tin of 60 °C, COVPmax increases from 3.95% to 4.71%, while at Tin of 90 °C, COVPmax reduces from 3.08% to 1.52%. Moreover, knock recognition achieves the best classification result at Tin of 60 °C. The average classification accuracy reaches 91.35%, and the best classification accuracy reaches 99.67%. This shows that feature extraction method based on EMD and SampEn has excellent classification performance for knock recognition of coal-based naphtha HCCI engines.