Experimental Study on Temperature Compensation for Dual-Axis MEMS Accelerometers Using Adaptive Mode Decomposition and Hybrid Convolutional–Recurrent Temporal Network Modeling

This paper presents a novel temperature compensation approach for dual-axis Micro-Electro-Mechanical System (MEMS) accelerometers, integrating Adaptive Mode Decomposition (AMD) with Grey Wolf Optimization (GWO) and Hybrid Convolutional-Recurrent Temporal Network (HCR-TN). The proposed method aims to address temperature-induced bias drift, which significantly affects accelerometer performance. Experiments were conducted across a temperature range from -40 °C to +60 °C to evaluate the effectiveness of the compensation algorithm. The results show considerable improvements in bias stability, with the compensation method successfully reducing temperature-induced drift across both axes. Additionally, the algorithm was tested under realistic conditions, including noise and mechanical disturbances, demonstrating its robustness in practical applications. These findings highlight the potential of the proposed method for enhancing the reliability and accuracy of MEMS accelerometers in real-world sensing environments.