【作者】 关广丰；

【导师】 李洪人；

【作者基本信息】 哈尔滨工业大学 ， 机械电子工程， 2007， 博士

【摘要】 本文以哈工大电液伺服仿真及试验系统研究所承接的国防863项目“三向六自由度液压振动试验系统”为背景,对六自由度液压振动试验系统的控制策略进行理论和实验研究。六自由度振动台控制系统由伺服控制系统和振动控制系统两部分组成。伺服控制系统的主要作用是将自由度驱动信号转化为各单系统的驱动信号,驱动平台运动,实现振动台的自由度控制。仅采用伺服控制方法可以控制平台运动,但此时系统的响应信号与输入的期望信号相比会有很大偏差。因此在伺服控制系统的外环加入振动控制系统,对驱动信号进行迭代补偿,使得系统的响应信号能够高精度地复现期望信号。振动台伺服控制系统主要由自由度合成及分解控制器、三状态控制器和压力镇定控制器三个部分组成。本文首先简述传统的六自由度控制策略,然后提出在原有六自由度控制回路中增设两个扭曲自由度的控制回路,采用所谓“八自由度控制策略”控制平台运动。采用六自由度控制策略时,驱动量的解不唯一,而且各激振器间内力受负载和油温等因素影响变化较大。采用八自由度控制方式时,驱动量的解唯一,各激振器间的内力变换较小,有利于提高系统的控制精度。接着对传统压力镇定控制器进行改进,将控制器的可调参数缩减为两个,且参数间相互独立,简化了调节过程。最后通过实验验证了八自由度控制策略及改进后压力镇定控制器的有效性。在振动台传统伺服控制系统中,自由度合成及分解矩阵都是基于零位线性化的假设得出的,而且在平台运动过程中矩阵元素值始终不变,控制算法不完善。本文对六自由度振动台进行运动学分析,给出基于运动学分析的振动台系统位姿控制结构,采用正解算法及雅可比阵分别代替传统伺服控制系统中的自由度合成和分解矩阵。仿真分析及实验研究表明,与传统矩阵控制方法相比,采用基于运动学分析的位姿控制策略能够有效改善振动台系统低频段的加速度均匀度,减小加速度横向分量。多自由度波形复现传统振动控制算法中大多采用H1法进行频响函数估计,而H1法假设系统输入端不存在噪声,其估计值为系统真实频响函数的欠估计。本文提出基于EV模型进行频响函数估计。在输入端和输出端均存在噪声的前提下,给出了基于EV模型的系统频响函数估计法。并根据多自由度波形复现试验系统的特点,对频响函数估计进行了简化。仿真分析及实验研究表明,在波形复现控制系统中应用基于EV模型的频响函数估计法可明显提高试验的控制精度。为进一步提高随机振动功率谱(PSD, power spectral density)复现的精度,在驱动谱迭代和驱动信号生成两个方面对传统功率谱迭代算法进行了改进。首先,提出将驱动谱迭代公式中的修正系数由实数变为实向量,在不同频段取不同修正系数,即采用分频段变步长法对驱动谱进行修正,提高了迭代算法的收敛速度。其次,提出利用驱动谱的信息采用Parks-McClellan法设计FIR滤波器,利用滤波器对白噪声信号进行滤波生成时域驱动信号,由于避免了传统控制算法中的时域随机化过程,缩短了算法的执行时间。仿真分析及实验研究表明,改进后的功率谱复现控制算法能够明显提高功率谱复现的精度。更多还原

【Abstract】 Upon the background of National Defense 863 Project to develop 3 axis 6 DOF (degree-of-freedom) hydraulic vibration test system, which is developed by IEST (Institute of Electro-hydraulic Servo Simulation & Test System), the control strategy of the vibration test system has been theoretically and experimentally studied.The control system of 6 DOF vibration table consists of servo control system and vibration control system. The servo control system is used to realize the DOF control mode, i.e., to convert the DOF drive signals to single actuator drive signals. There have large deviations between response signals and input signals with the only servo control system. So the vibration control system is added in the outer control loop of the servo control system. Based on the FRF (frequency response function) of the system, the drive signals are corrected so that the response signals can replicate the expect signals in high degree of accuracy.The servo control system of vibration table consists of DOF composition and decomposition controller, three variable controller and force balance controller. Based on the analysis of 6 DOF matrix control method, 8 DOF matrix control method is presented in this dissertation. Two torsion DOF control loop are added to the original 6 DOF control loop. As the existence of no uniqueness solutions of actuator drive signals in the 6 DOF control mode, the inner forces among actuators vary within wide limits. In 8 DOF control mode, the solutions of actuator drive signals are uniqueness, so the inner forces among actuators are almost constant. The const inner forces are favorable to improve the control precision. Then, the conventional force balance controller is improved. The number of adjustable parameters is reduced to two. The two parameters are independent and can be adjusted individually. At last, experimental results validated the 8 DOF matrix control method and the improved force balance controller.The DOF composition and decomposition matrixes in classical servo control system are got based on the assumption of linearization near the zero position. The matrixes are invariant. This indicates that the conventional matrix control method is inaccurate. Kinematics analysis and the pose control structure of 6 DOF vibration table are presented in this dissertation. Forward solution and jacobian matrix replace the DOF composition and decomposition matrixes in the conventional servo control system. Simulation and experimental results show that the pose control structure based on the kinematics analysis of 6 DOF vibration table is favorable to improve the performance indices of acceleration evenness degree and transverse components in the low-frequency band.FRF H1 estimator is used in the classical 6 DOF waveform replication algorithm to estimate the FRF of the system. It’s well known that the H1 estimator is influenced by the noise in the input signals and generates an under-estimation of the true FRF. This paper presents the FRF estimator based on the EV model to reduce the bias error of FRF H1 estimator. The FRF estimator based on the EV model takes into account the errors in both the inputs and outputs of the system and would lead to more accurate FRF estimation. Based on the features of MIMO waveform replication control system, the estimator is simplified. The results of simulation and experiment show that the waveform replication algorithm with the FRF estimator based on the EV model can improve further the control precision.In order to improve the control precision of random vibration PSD replication, the correction algorithm of drive signals PSD and the generation of time domain drive signals are improved in the classical algorithm. First, the drive signals PSD iteration algorithm with different step length in different frequency band is presented to raise the rate of convergence. Second, the method of generating the drive signal in time domain by filtering a series of independent white noise with designed FIR filter is presented. The FIR filter is designed by the Parks-McClellan algorithm with the information contained in the in the drive signal PSD. Due to avoiding the time domain randomization procedure, the execution time of the algorithm is reduced. The results of simulation and experiment show that the improved PSD iteration algorithm is favorable to improve the control precision.更多还原