Temperature Compensation Method Based on an Improved Firefly Algorithm Optimized Backpropagation Neural Network for...

Area of Science:

• * Instrumentation and Measurement
• * Artificial Intelligence in Engineering
• * Materials Science and Engineering

Background:

• * Micromachined Silicon Resonant Accelerometers (MSRA) exhibit significant output drift in variable temperature environments.
• * Temperature-induced errors degrade accelerometer performance, necessitating effective compensation strategies.

Purpose of the Study:

• * To develop and validate an improved firefly algorithm-backpropagation (IFA-BP) neural network for temperature compensation of MSRA.
• * To enhance the accuracy and robustness of temperature compensation compared to existing methods.

Main Methods:

• * Development of an Improved Firefly Algorithm (IFA) to optimize the initial weights and thresholds of a Backpropagation (BP) neural network.
• * Training and evaluation of the IFA-BP model using zero-bias and full-temperature experimental data from an MSRA.
• * Comparative analysis against the standard Firefly Algorithm-Backpropagation (FA-BP) model.

Main Results:

• * The IFA-BP model demonstrated superior temperature compensation performance over the FA-BP model.
• * Zero-bias stability at room temperature improved by over an order of magnitude post-compensation.
• * In the -40°C to 60°C range, scale factor variation improved >70 times, and bias variation improved ~3 orders of magnitude.

Conclusions:

• * The proposed IFA-BP neural network effectively compensates for temperature-induced drift in MSRAs.
• * This method significantly enhances accelerometer stability and accuracy across a wide temperature range.
• * The IFA-BP approach offers a robust solution for improving accelerometer performance in demanding environments.