Recent Advances in Improving the Thermo-Electro-Mechanical Responses of MEMS Resonant Accelerometers via a Novel Bidirectional Long Short-Term Memory: A Systematic Review

Abstract

Microelectromechanical systems (MEMS) resonant accelerometers have gained significant attention due to their high precision, stability, and suitability for applications in aerospace, navigation, and structural health monitoring. However, their performance is often degraded by thermo-electro-mechanical (TEM) coupling effects, including temperature-induced drift, electrical noise, and mechanical nonlinearities. Recent advancements in data-driven approaches, particularly deep learning, have shown promise in mitigating these challenges. This systematic review explores recent developments in improving TEM responses of MEMS resonant accelerometers through the integration of Bidirectional Long Short-Term Memory (BiLSTM) networks. The study critically analyzes existing methodologies that combine physical modeling with machine learning techniques to enhance signal compensation, reduce drift, and improve accuracy. Emphasis is placed on hybrid frameworks that utilize temporal sequence learning to capture bidirectional dependencies in sensor data. The review also highlights key trends in preprocessing techniques, feature extraction strategies, and model optimization approaches. Comparative insights are provided on performance metrics such as sensitivity, stability, and noise reduction. Furthermore, the paper identifies research gaps and future directions, including real-time deployment and energy-efficient implementations. The findings demonstrate that BiLSTM-based approaches significantly outperform traditional compensation techniques, offering robust solutions for next-generation MEMS accelerometer systems.