Research on Temperature Drift Compensation of Inclination Sensor Based on Improved BP Neural Network

Abstract

Inclination sensor is easily influenced by the temperature to produce the phenomenon of temperature drift, which affects measurement accuracy. In order to improve the accuracy of inclination sensor and overcome the shortcomings of slow convergence speed and low accuracy of BP neural network, an improved genetic simulated annealing algorithm was proposed to optimize the BP neural network to temperature drift compensation of the inclination sensor. In this algorithm, a sort reservation and elimination strategy is proposed to improve the convergence speed of the network. In order to improve the compensation accuracy of the network, the relationship between iteration times, population fitness, crossover probability and mutation probability is considered in the algorithm iteration process. In the experiment, temperature and angle are used as input nodes, and the compensated angle is used as output node to build a network model. Compared with the experimental results of BP and GASABP algorithms, the algorithm has fast convergence speed and high compensation accuracy. The average temperature drift error after compensation is 0.028, which is an order of magnitude higher than the maximum temperature drift error of 0.4 before compensation.