Artificial Intelligence Techniques for Improving the Thermo-Electro-Mechanical Responses of MEMS Resonant Accelerometers via a Novel Bidirectional Long Short-Term Memory: Trends and Challenges

Abstract

Microelectromechanical systems (MEMS) resonant accelerometers have emerged as critical components in high-precision sensing applications, including aerospace navigation, structural health monitoring, and consumer electronics. However, their performance is significantly influenced by thermo-electro-mechanical (TEM) coupling effects, leading to drift, nonlinearity, and reduced sensitivity. Recent advancements in artificial intelligence (AI), particularly deep learning, have introduced novel approaches to model and compensate for such complex interactions. This study presents a comprehensive review of AI-driven techniques, with a particular focus on Bidirectional Long Short-Term Memory (BiLSTM) networks, for enhancing the performance of MEMS resonant accelerometers. The paper explores how BiLSTM models effectively capture temporal dependencies in sensor data affected by temperature fluctuations, electrical noise, and mechanical disturbances. Furthermore, it examines optimization strategies, hybrid architectures, and data-driven compensation methods that improve accuracy and stability. Emerging trends such as edge AI deployment, physics-informed learning, and real-time adaptive calibration are also discussed. Despite significant progress, challenges remain in terms of dataset availability, model interpretability, computational constraints, and robustness under extreme conditions. This review consolidates existing research, identifies gaps, and provides future directions for integrating AI techniques into next-generation MEMS accelerometer systems, enabling enhanced reliability and precision in dynamic environments.