A Random Error Suppression Method Based on IGWPSO-ELM for Micromachined Silicon Resonant Accelerometers.

Area of Science:

• Instrumentation
• Signal Processing
• Machine Learning

Background:

• Micromachined silicon resonant accelerometers (MSRA) suffer from complex and uncertain random errors, limiting their practical application accuracy.
• Existing methods for error suppression in accelerometers often face challenges in effectively handling complex noise compositions.

Purpose of the Study:

• To propose and validate a novel random error suppression method for MSRA.
• To enhance the output accuracy of MSRA by addressing complex random error components.

Main Methods:

• A modified wavelet threshold function was employed for white noise separation from the useful signal.
• An improved grey wolf and particle swarm optimized extreme learning machine (IGWPSO-ELM) was developed, incorporating non-linearized factors for enhanced convergence.
• A three-dimensional input was constructed using previous and current output frequencies.

Main Results:

• The proposed IGWPSO-ELM method effectively reduced velocity random walk from 4.3618 μg/√Hz to 2.1807 μg/√Hz.
• Bias instability was decreased from 2.0248 μg to 1.3815 μg.
• Acceleration random walk was suppressed from 0.53429 μg·√Hz to 0.43804 μg·√Hz.

Conclusions:

• The IGWPSO-ELM method demonstrates significant effectiveness in suppressing random errors in MSRA.
• This approach leads to a substantial improvement in the overall accuracy and reliability of micromachined silicon resonant accelerometers.