Hybrid Filtering Compensation Algorithm for Suppressing Random Errors in MEMS Arrays.

Area of Science:

• Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Microelectromechanical systems (MEMS) often suffer from high error rates due to poor signal-to-noise ratios.
• Random errors in MEMS sensor arrays significantly impact their accuracy and reliability.

Purpose of the Study:

• To propose and validate an online compensation algorithm for mitigating random errors in MEMS arrays.
• To enhance the signal-to-noise ratio and overall performance of MEMS devices.

Main Methods:

• Development of a wavelet threshold back-propagation neural network (WT-BPNN) algorithm.
• Integration of denoising and error compensation using a back propagation neural network (BPNN).
• Implementation and testing of the algorithm on a ZYNQ-based MEMS array hardware platform.

Main Results:

• Significant improvements in gyroscope performance: 12 dB reduction in zero-bias instability, 10 dB in angular random wander, and 7 dB in angular velocity random wander in static conditions.
• An 8 dB reduction in output data dispersion across various dynamic environments.
• Experimental validation of the algorithm's robustness and feasibility.

Conclusions:

• The proposed WT-BPNN algorithm effectively suppresses random errors in MEMS arrays.
• The algorithm demonstrates significant performance enhancements for MEMS sensors, particularly gyroscopes.
• The study confirms the robustness and practical applicability of the WT-BPNN for MEMS error compensation.