Engine misfire is a combustion fault that results in one or more cylinders not contributing to the torque. It negatively affects the drivability, consumes more fuel, produces higher emissions and can cause damage to exhaust after-treatment systems if not detected. Conventional on-board diagnostics are primarily based on crankshaft speed variation, which is useful for regulatory monitoring, but can be less useful in retrofit maintenance systems, older engines and small marine engines that don’t have advanced diagnostics. This review considers low cost, non-invasive misfire detection approaches based upon vibration and acoustic sensing, focusing on real vehicle and small marine applications. The paper compares crank-speed, exhaust-temperature, vibration, acoustic, acoustic-emission, and sensor-fusion methods, considers signal-processing and learningbased approaches, and addresses issues of embedded feasibility, noise robustness and sensor placement, data acquisition. The literature review reveals a high level of accuracy in the laboratory, but limited applicability of the practical deployment due to road vibration, ambient sound, mount repeatability, engine-to-engine variability and limited public datasets. Recent studies on acoustic-vibration fusion, frequency domain method and spectral analysis of marine diesel show the promising trend for low-cost embedded prototypes. The most practical nearterm solution for practical application is a lightweight embedded system that includes an engine mounted vibration sensor, a body mounted reference vibration sensor, a digital microphone, and then windowed RMS/spectral features and transparent decision logic. The review ends by identifying gaps in the research and designing recommendations to build a strong and usable prototype that can alert operators at an early stage without relying on vehicle OBD infrastructure.

A REVIEW OF LOW-COST EMBEDDED ACOUSTIC AND VIBRATION-BASED ENGINE MISFIRE DETECTION FOR REAL-VEHICLE