Temperature Modeling Study for High Precision Gyroscope Based on Neural Network

Abstract

In the study, neural network theory was used to build a nonlinear model for high precision gyroscope reflecting the relationship between temperature and drift.The result shows that types of neural network and input sample have great influence on model precision. High precision gyroscope is sensitive to temperature. The input sample must take account of the continuous temperature and mean temperature value in a period of time can not be used for model. The model of multi-input and single-output is better than the model of single-input and single-output in the same neural network. Genetic algorithm(GA) can optimizes Back-Propagation(BP) neural network. GA-BP and BP neural network can’t achieve the precision request. Radial basis function(RBF) neural network has good precision whose relative error is about 10-6. RBF neural network can achieve model request.