Advancements in detecting combustion events through vibration analysis in internal combustion engines: a literature review

Abstract

Accurate combustion characterization is very important for the reliability, efficiency, and emissions of internal combustion engines (ICEs). Standard in-cylinder pressure sensors are costly, obtrusive, and have a short lifespan under harsh environments. This research examines vibration-based analysis as a non-intrusive method for identifying and reconstructing combustion events in compression ignition (CI) and homogeneous charge compression ignition (HCCI) engines. We examine the correlation between engine block vibrations and combustion parameters, including start of combustion (SOC), peak pressure (PP), pressure rise rate (PPRR), maximum heat release rate (HRRmax), and mass fraction burned (MFB), while assessing signal decomposition techniques such as Empirical Mode Decomposition (EMD) and Variational Mode Decomposition (VMD). Advanced signal processing is also combined with machine learning (ML) models to improve real-time estimation and nonlinear mapping. Our study shows that vibration signals can be used as reliable indicators of combustion, with timing precision of less than 1 CAD and pressure reconstruction error of less than 5%. Nonetheless, difficulties persist in adaptive parameter tuning, signal decoupling during transient situations, and the comprehensibility of machine learning models. Hybrid and explainable ML frameworks, multi-sensor data fusion, and the creation of virtual combustion noise sensors for onboard diagnostics are some promising areas of research. In the end, vibration-based combustion analysis provides a cheap and scalable means to get to real-time, non-intrusive monitoring, which will help the shift to cleaner, quieter, and more efficient engines.