Deep Learning and Optimization Approaches in Enhancing Thermo-Electro-Mechanical Responses of MEMS Resonant Accelerometers with an Attention-Guided Siamese Fusion Neural Network: A Review

Abstract

Microelectromechanical Systems (MEMS) resonant accelerometers are widely employed in aerospace, automotive, structural health monitoring, biomedical devices, and inertial navigation due to their high sensitivity and stability. However, their performance is significantly affected by thermo-electro-mechanical coupling, including temperature-induced stress, electrostatic nonlinearity, thermoelastic damping, and structural vibrations, which reduce measurement accuracy and long-term reliability. Conventional analytical and finite element models often struggle to capture these complex nonlinear interactions under varying operating conditions. This review examines recent advances in deep learning and optimization techniques for improving MEMS resonant accelerometer performance, with particular emphasis on the Attention-Guided Siamese Fusion Neural Network (AGSFNN). The AGSFNN framework integrates dual-stream feature learning with attention mechanisms to effectively fuse thermal and mechanical information, enabling accurate prediction and compensation of resonant frequency variations. The review also analyzes complementary approaches, including Convolutional Neural Networks, Long Short-Term Memory networks, Graph Neural Networks, Physics-Informed Neural Networks, transformer-based architectures, and optimization algorithms for thermal compensation, fault diagnosis, and nonlinear response modeling. Overall, the study highlights intelligent neural network-based compensation as a promising direction for developing robust, self-adaptive, and high-precision MEMS resonant accelerometer systems for next-generation sensing applications.