【作者】 王威立；

【导师】 陈娟；

【作者基本信息】 中国科学院研究生院（长春光学精密机械与物理研究所） ， 机械电子工程， 2012， 博士

【摘要】 随着科学技术的发展，目标的机动性越来越强，对光电跟踪系统的要求也越来越高。转台伺服系统作为光电精密跟踪系统的硬件设备，在光电跟踪系统的研制中起着极其重要的作用。转台伺服系统的精度主要受伺服控制器和控制方法的影响，因此选择高性能控制器和采用先进的控制算法是提高跟踪精度的主要途径。本文基于Compax3伺服控制器建立了光电跟踪伺服平台，并对共轴跟踪技术进行了探讨和实验验证。首先介绍了光电跟踪系统的概念及组成，并对光电跟踪系统中的现有伺服控制算法和控制器的发展现状进行了简要总结。介绍了复合控制、等效复合控制和共轴跟踪原理。实现共轴跟踪的关键在于获得目标位置、速度和加速度信号。采用预测滤波技术可以得到目标信息，但是随着跟踪目标的多样性及机动性越来越强，难以找到一个运动模型去适合所有的观测对象。本文提出利用ELM神经网络对脱靶量，伺服转台位置、速度和加速度进行数据融合得到目标位置、速度及加速度。针对ELM神经网络运算量大，对ELM神经网络算法进行优化，缩短了运算时间，运算时间大约为4.58ms，达到了光电跟踪系统的实时性要求。并根据某光电跟踪系统的实验数据进行了ELM神经网络数据融合仿真。针对实现共轴跟踪需要的信息源，在伺服转台上安装了角加速度传感器。分析了角加速度传感器的工作原理和模型，利用频率测试法得到角加速度传感器的传递函数，并进行滞后补偿。通过实验比较了在角加速度传感器和光电位置编码器两种方式下，分别得到的伺服转台的速度和加速度信号的优劣。最后，基于Compax3伺服控制器建立了光电跟踪伺服平台，对平台的主要参数进行性能测试。用程序模拟了光电探测器的特性，并进行了共轴跟踪实验验证。当目标运动最大速度为50°/s，最大加速度为30°/s~2时，系统最大跟踪误差由简单闭环控制时10.35′减小到0.38′。实验结果表明本论文所采用的方法具有更高的实时性和精确度，能有效提高系统的跟踪精度。更多还原

【Abstract】 The target mobility is increasingly strong with the development of science andtechnology, so the requirements for the electro-optical tracking system performancebecome higher and higher. The turntable servo system, as the hardware ofelectro-optical tracking system, is extremely important. The tracking precision ofservo system depends on servo controller and servo control method, so the mainmeasures for improving the tracking precision are using higher performancecontroller and more advanced control method. In this paper the electro-opticaltracking servo turntable is set up based on Compax3servo controller, furthermore,the on-axis technology is discussed and verified by experiment.Firstly, the concept and component parts of electro-optical tracking system areintroduced, the current developing situation of servo controller and method are alsobriefly summarized.The principles of compound control, equivalent compound control and on-axisare introduced, respectively. The key point to realize the on-axis tracking is how toprecisely capture the position, velocity and acceleration of the target. Although thetarget information can be predicted through the filter and prediction, it is impossibleto get a general model which can be applied to all objects because of the complexdiversity and increasing mobility of the tracked targets.. In this paper the nerve netextreme learning machine (ELM) was adopted to obtain the target motioninformation through data fusing between the target miss distance and the movingstate of turntable. The ELM neural network algorithm is optimized in this study to reduce thepervious larger amounts of computation. The period of ELM is decreased to4.58msand meets the real-time requirement of the electro-optical tracking system The ELMsimulation is done according to actual experimental data sampled from anelectro-optical tracking system.In order to obtain the turntable acceleration, the acceleration sensor is fixed onthe turntable. The test principle and model of acceleration sensor are analyzed. Andthe transfer function of acceleration sensor is acquired through the freq uency methodand furthermore regulated. The turntable velocity and acceleration are contrastedwhich are obtained separately by encoder and acceleration sensor.The turntablevelocity and acceleration obtained by encoder and acceleration sensor, respective ly,are contrastedFinally, the electro-optical tracking servo platform is established based on theCompax3servo controller, and the main parameters and performance of the platformare tested. The feature of detector is simulated by programmer, and theelectro-optical tracking experiment is done. When the target maximal velocity is30/sand accelerator25/s2, the maximal system tracking error decreased from10.35′, when the system was a closed-loop control, to0.38′when that of on-axis.This paper mainly experimentally verified the target motion information, gotten bythe novel method, had better real time performance and higher accuracy, which isexpected to be a potential to improve the system tracking precision significantly.更多还原