The Research Of Control Method For Electric Powered Servo Load System

Servo load system, which is an important equipment in hardware-in-loop simulation, provides ground experimental data for studying control performance of aerocraft, and also offers reliable testing information for application of new technique. Servo load system has being an emphasis on national defence, aviation, spaceflight field and other science research, industry production since about twenty years ago. This thesis analyses and studies the electric powered servo load system deeper based on the load system of one minitype robot bomb undercarriage, and brings forward some effectual control methods.To restrain additional torque is the key in servo load system. In this thesis, firstly, the holistic system mathematic model is established, the cause of additional torque is analysed, then conventional control methods-the angle displacement feedforward compensation and the angle velocity of load motor inner feedback compensation are introduced briefly, based on that, the current of load motor inner feedback compensation is brought forward, it reduces orders of differential coefficient in compensation part, then reduces the difficulty of realizing compensation part greatly.Secondly, considering that, compensation methods upwards mostly are researched with an eye to torque load system, and displacement servo system always is ignored, so big limit exists. This thesis based on structure-invariable theory brings forward the displacement servo motor current feedforward compensation for the first time, which overcomes the limit of the methods upwards, and good compensation effect is found from theoretic simulation in linear constant system.Finally, in view of that, the nonlinearity (such as friction nonlinearity, clearance nonlinearity in mechanism connection) and parameter variety can not be solved by conventional control methods, a novel complex control method is brought forward based on neural network generalized predictive control theory, detailed control structure and algorithm is shown in chapter 4. Briefly complex control is that, a complex neural network is used for on-line system identification, and a generalized predictive controller makes use of the result from complex neural network identifier, roll- optimizes, controls the system. Simulation is done in Matlab, we can see that, for stribeck-friction and parameter variety, the complex controller has stronger adaptive capacity, puts up stronger robustness, improves the dynamic loading performance greatly.