Temperature drift modeling and compensation of capacitive accelerometer based on AGA-BP neural network

Abstract

Abstract The accuracy is one of the key indicators for capacitive accelerometers. However, it will be affected by temperature drift especially the rapid temperature variation. To compensate the temperature drift, a back propagation (BP) neural network model based on an adaptive genetic algorithm (AGA) has been proposed. According to the ratio of average fitness to maximum fitness, a nonlinear equation constructed is to obtain adaptive probabilities of crossover and mutation. Moreover, the algorithm flow of crossover and mutation is also adaptive by this ratio. Hence, the compensation model will avoid local optimal values effectively. Five experiments with different input accelerations have been conducted in a temperature variation rate of 1 ℃/min to test the model. The results validate that, the AGA-BP compensation model has the best effect, compared to the multiple linear regression (MLR) compensation model and GA-BP compensation model. This model is of high reference values for compensating rapid temperature variation.