Development of a Vibration-Based System to Detect Internal Combustion Engine Faults

Abstract

Engine misfire in internal combustion engines (ICEs) is a critical fault condition that leads to increased emissions, reduced fuel efficiency, and accelerated mechanical degradation. While conventional on-board diagnostic systems are capable of detecting misfire events, they lack the capability to spatially localize the fault source within the engine structure of an automobile. This paper presents the development and real-world validation of a low-cost vibration-based system for accurate detection of the spatial coordinates of the misfiring cylinder (localization). The proposed system utilizes four ADXL345 three-axis accelerometers coupled with ESP32 microcontrollers, transmitting real-time sensor data via a cloud-based pipeline to a web application for user interactive fault visualization. A signal processing pipeline containing sensor calibration, Fast Fourier Transform (FFT) and Short-Time Fourier Transform (STFT) frequency analysis, and Butterworth band pass filtering was developed and implemented in order to isolate fault-relevant vibration signatures. Received Signal Strength (RSS)-based localization combined with trilateration and multi-lateration geometric solving algorithms was employed to estimate the spatial location of the vibration source in both the two-dimensional and three-dimensional configurations. The system was first validated on a controlled experimental rig and then deployed across three real-world engine scenarios with deliberately induced misfire conditions. Experimental results demonstrated successful misfire cylinder identification in all these test cases, with a maximum localization error of 8%. The proposed framework offers a scalable and cost-effective approach to existing diagnostic approaches, with further potential extensions to rotating machinery fault diagnosis, marine engine health monitoring, and autonomous vehicle propulsion system diagnostics.