Study On Neural Network Identification Of Rotary Stabilized Platform Control

Rotary steerable drilling technology, which can improve the oil and gas production and recovery, is a full newly drilled well technology. A lot of equipments have been developed and monoplized of overseas firms. The rotary drilling tools are exploited in China.The structures of the rotary steerable drilling system and also of stable platform structure, characteristic and principle of operation are analyzed. Lugre friction model is used to describe friction phenomena of the stable platform.Aiming at the shortcoming of slow convergence, an improved BP algorithm is put forward. It combines an excitation function with a sharpness factotial and joins coordinator in the error reversion transmission which complete transmission changes into section transmission. The simulation results show that the study speed is higher than classic BP algorithm.A neural network parameter identification method is presented, basing on weight value boundary problem. The simulation results indicate that the neural network method is fitted for lugre model parameter identification.And the neural network PID is designed for the stable platform which contains lugre model. Simulation results show that the self-adaptability and anti-jamming ability are improved with neural network PID control. At the same time, it has been reviewed that each parametric change of the lugre model for system performance'impact, these are useful to improve performance of stabilized platform control system.