神经网络在自动调平系统中的应用

Neural Network Application in Automatic Leveling System

针对雷达天线车调平系统的调平控制问题,提出了刚性模型与神经网络误差补偿模型相结合的控制结构。通过纯刚性系统模型假设和对平台姿态与各支腿关系的分析,提出了固定高度过调修正的调平策略,并利用逆向建模法建立了纯刚性系统的自动调平数学模型。由于难以对某些重大影响因素建立精确的数学模型,为提高调平精度,利用神经网络对未知函数的辨识能力建立了神经网络误差补偿模型。平台在横向倾角、纵向倾角的初始值分别为1°和-1.5°时,只经二次预测调整,即可使θ_x收敛在0.005°,θ_y收敛在-0.005°。仿真结果表明,该控制结构具有较高的控制精度、较快的收敛速度。

According to the control problem of horizontal adjustment system of a radar-antenna truck, the control structure composed by rigidity model and neural network error compensation model is proposed. Based on the supposition of leveling system being rigidity system and the analysis of the relationship between the pose of platform and the length of legs, the fixation height over adjusting correction leveling strategy is proposed, and the automatic leveling model of rigidity system is founded by using converse modeling method. The neural network error compensated model is founded by means of the ability to distinguish an unknown funetion. When the initial level obliquity of transverse and lengthways are 1° and - 1.5° respectively, the θx converges at 0.005°, θy converges at - 0.005° only by 2 times of adjustments. Simulation result shows that the control structure has high control precision and rapid convergence ability.