A Low-Cost Broadband MEMS Vibration Sensing System with Encapsulated Mounting and Embedded Spectral Feature Extraction for Automotive Applications

Low-cost MEMS accelerometers are increasingly used for vibration-based condition monitoring, but their high-frequency performance is often limited by the way they are mechanically mounted. In many practical implementations, the coupling between the sensor and the monitored structure is poorly controlled or only marginally addressed, leading to mounting-induced resonances that can compromise broadband measurements. This paper presents a low-cost wideband vibration sensing system based on a MEMS accelerometer and an epoxy-encapsulated mounting solution specifically designed to improve mechanical coupling and shift parasitic resonances toward higher frequencies. Unlike approaches that mainly focus on signal processing, the proposed system experimentally addresses the mechanical integration of the sensor as a key requirement for reliable vibration measurements. The sensing unit is combined with embedded spectral feature extraction, which reduces data dimensionality by retaining the most relevant frequency components and provides compact feature vectors suitable for Artificial Intelligence (AI) and Machine Learning (ML) based condition monitoring. Finite element simulations, laboratory frequency-response tests, and automotive measurements on different engine configurations show that the proposed approach improves the dynamic behavior of the sensor and captures distinctive vibration signatures for automotive and industrial diagnostic applications.