BO-LSTM-Based TDE Precise Estimation Model of Capacitive MEMS-Gyros Using Thermal-Induced Physical Characteristics...

Area of Science:

• * Engineering and Materials Science
• * Sensor Technology and MEMS Devices

Background:

• * Conventional models for Capacitive MEMS-Gyros (CMG) Temperature Drift Errors (TDE) suffer from incomplete Temperature Correlated Quantities (TCQ) and suboptimal parameter identification, limiting bias stability.
• * Thermal-induced physical variations in CMG and Silicon-based materials affect stiffness and introduce TDE, necessitating a more robust estimation approach.

Purpose of the Study:

• * To propose a novel BO-LSTM-based TDE precise estimation model for CMGs.
• * To improve the bias stability and environmental adaptability of CMGs by accurately estimating TDE.

Main Methods:

• * Analysis of thermal-induced microstructural deformation and stiffness deterioration to identify complete TCQ, including ambient temperature variations and their higher orders.
• * Application of Long Short-Term Memory (LSTM) for parameter identification of the modified TDE model, leveraging time series characteristics.
• * Integration of Bayesian Optimization (BO) to optimize LSTM hyperparameters for enhanced performance.

Main Results:

• * The BO-LSTM model accurately estimates TDE in CMGs, outperforming conventional Radial Basis Function Neural Network (RBFNN) models.
• * Bias stability of the CMG was improved by 20% using the proposed model.
• * The model effectively decouples temperature dependence and enhances environmental adaptability.

Conclusions:

• * The BO-LSTM model offers a significant advancement in TDE estimation for CMGs.
• * Improved bias stability and environmental adaptability make CMGs more reliable in diverse conditions.
• * This approach provides a robust solution for mitigating temperature-related performance degradation in MEMS gyroscopes.