Self-Detection of Mounting Looseness with a MEMS Accelerometer

MEMS accelerometers play an increasingly critical role in condition monitoring of machines and infrastructure. Combined with 3D printing technologies, the rapid prototyping of smart sensors enables the exploration of novel sensing technologies. However, their performance critically depends on proper mounting, which affects the reliablity of vibration data. This study investigates the potential of MEMS accelerometers to assess their own mounting integrity and whether this capability extends to various materials used for 3D printed sensor mountings. To investigate this, a sensor fixture was 3D printed for four commonly used materials and the elastic Young’s modulus of each one was experimentally determined. A vibration analysis was conducted by exciting a securely mounted adapter and measuring its response with a 3D laser doppler vibrometer. Finite element simulations validated that the equivalent modulus reduced deviations between simulations and measurements. To explore the impact of loose mounting, laser measurements were taken under varying looseness conditions. After integrating an ADcmXL3021 sensor, extreme cases of looseness were introduced by removing bolts from the adapters fixture and exciting it with white noise. From resulting data, key vibration features were extracted and used to train Support Vector Machine classifiers: one using the full set of key features and one using only the resonance frequency. Both methods successfully distinguished securely mounted from loosely mounted conditions, though the singlefeature classifier showed reduced accuracy when generalizing across different materials. This self-supervising approach provides insights into the dynamics of mounting looseness and promises enhanced data integrity for MEMS accelerometers with 3D printed adapters.