The Modueling And Analysis Of Laser Gyro Temperature Error Based On Neural Network

Laser gyro is an indispensable component in inertia navigation system. Its precision and startup error will directly affect the precision and time of navigation. Errors of laser gyro can be classified as constant drift and random drift, for which mature methods can be applied to compensate. On account of inner conformation's particularity and absence of temperature control device, comparing with gyro that have temperature device, temperature has more prominent impacts on the precision of laser gyro. So we need to build up a temperature error model aiming at compensating for the unfavorable results caused by the change of temperature. This thesis focuses on research as followed:It's elaborated the temperature model compensation theory and the practical significance. From establishment model correlation theories knowledge obtaining, it's introduced the laser gyroscope principle of work, the laser gyroscope basic error and the effects of temperature to laser gyro.Compares with the traditional mathematics model, this article has used the neural network modeling. It only needs through the input and output to recognize the mapping relations, and approaches the input and output data through the training network self-adjusting network weight, avoided determining the coefficient through the solution system of equations the work. It has also made the detailed correlation introduction to the neural network aspect article, for example: Neuron structure, transfer function, neuron network, study rule, BP network and RBF network structure and their training algorithm and so on.In modeling and simulation analysis, according to laser gyroscope principle equation development, neural network model principle formula, and explained in the formula the parameter selection, the test, processing. First from which is affected the temperature bias carries on to the input temperature and the output the pretreatment, it provides the reliable basis for the modeling. In the experimental simulation uses the most typical BP network and the radial direction base network took the modeling network, separately after the network design, the network training, the parameter hypothesis network test, the simulation result and so on several aspects makes the temperature compensation...