Laser gyro temperature compensation using modified RBFNN.

Area of Science:

• * Instrumentation and Measurement Science
• * Artificial Intelligence and Machine Learning

Background:

• * Laser gyros are susceptible to temperature variations, affecting their zero bias and output stability.
• * Traditional compensation methods like multiple linear regression (MLR) and standard radial basis function neural networks (RBFNN) have limitations in accuracy and computational efficiency.

Purpose of the Study:

• * To develop an improved RBFNN for accurate compensation of laser gyro zero bias under varying temperatures.
• * To enhance the stability and reliability of laser gyro output through advanced neural network techniques.

Main Methods:

• * A modified RBFNN integrating a Kohonen network for pattern classification and an orthogonal least squares (OLS) algorithm for optimal center selection.
• * Comparative analysis using MLR, standard RBFNN, and the modified RBFNN on gyro data under diverse temperature conditions.

Main Results:

• * The modified RBFNN effectively overcomes the limitations of randomly selected centers in traditional RBFNN.
• * Achieved compensated gyro output accuracy of 7.0 × 10-4 °/h, outperforming standard RBFNN (9.0 × 10-4 °/h) and MLR (1.4 × 10-3 °/h).
• * Reduced computation time by approximately 60 seconds compared to the traditional RBFNN.

Conclusions:

• * The proposed modified RBFNN offers a superior solution for stabilizing laser gyro output by accurately compensating for temperature-induced zero bias.
• * The method provides enhanced accuracy and computational efficiency, making it suitable for real-time applications.