【作者】 刘达；

【导师】 李木国；

【作者基本信息】 大连理工大学 ， 控制理论与控制工程， 2013， 博士

【摘要】 随着人类对海洋资源开发的深入,提高对海洋认知程度的要求也愈加迫切,作为海洋科学实验中进行波浪模拟的重要设备——造波机系统,也在研究中扮演着越来越重要的角色。造波机系统是在波浪理论、自动控制、机械工程等多学科交叉基础上发展起来的大型自动化设备,而以自动控制为基本原理的运动控制系统在造波机系统构建中起着核心的关键作用,决定着造波机系统性能的好坏。此外,网络化运动控制系统的出现,给造波机系统研发带来新机遇的同时也使之在技术上面临着新的挑战。本文正是在此背景下以研发高性能网络化造波机系统为目标,对网络化运动控制系统进行了深入研究,具体内容包括：首先,根据实际物理系统建立了造波机系统的运动控制模型,分析了系统负载及转动惯量特性。结合矢量控制和空间矢量脉宽调制技术提出一种基于小波神经网络的自适应PID控制算法,克服了传统PID控制在应用于具有非线性、强耦合、多变量的运动控制系统时遇到的困难,充分利用了小波神经网络良好的函数逼近能力和自适应能力,为解决复杂的非线性、不确定系统的控制问题提供了新思路。为了提高位置伺服系统的跟踪精度,提出基于小波神经网络自适应反推控制策略,该方法同时考虑位置与电流的跟踪性能,在保证系统状态变量全局渐近稳定前提下,获得了良好的动静态效果和鲁棒性。在此基础上,为进一步提高系统的抗扰能力,在设计中加入滑模控制,同时采用新的小波神经网络结构加速系统收敛,得到自适应滑模反推控制策略,仿真结果证明了该策略在跟踪性能和鲁棒性能上都具有优越的特性。其次,对引入网络的造波机运动控制系统进行了分析,并建立了造波机网络化运动控制系统的单轴和多轴模型。针对在模型分析中网络延时给系统带来的问题,提出用小波神经网络对工业以太网络延时进行预测,根据输入的过去时间延迟序列预测输出下一采样时刻的网络延时值。小波神经网络的参数通过训练算法实时更新,以保证预测输出的准确性。经过对实际以太网延时数据样本的预测分析表明该预测模型能够有效估计延时,具有实际的应用价值。对于运动控制系统中存在的时延效应进行了分析,考虑利用Smith预估器将系统中的纯延时环节移出闭环以保证系统稳定性,为克服Smith预估器对于模型敏感的缺点,提出一种带扰动补偿的自适应Smith预测控制策略,利用小波神经网络自适应调整参考模型使之输出特性趋于实际系统,同时利用小波神经网络对影响系统鲁棒性的扰动进行前馈补偿,使系统对扰动不敏感。仿真结果表明提出的控制策略可有效抑制扰动,具有良好的鲁棒性和动静态品质。为满足造波机运动控制系统的多轴同步控制要求,提出了相邻耦合滑模同步控制策略,将跟踪误差和同步误差同时考虑并加以消除,控制器的设计考虑了系统的不确定性,采用滑模控制方法加以抑制,使系统获得了良好的位置伺服和同步控制性能。再次,在EtherCAT标准工业以太网协议基础上搭建了网络化的运动控制系统,完成了从站的网络控制器及DSP控制器的软硬件设计。该系统通过主站利用实时工控软件由从站网络控制器与DSP控制器通信,DSP控制器在完成与网络控制器通信的同时还负责运动控制算法的实现。上位机发送的控制命令经由网络传输至控制器后,控制器根据其内部的模糊PID控制算法生成驱动信号,经差分转换电路传输至驱动器从而控制电机的运动,实现了的网络化运动控制。实验结果表明该网络化运动控制系统性能良好,同时证明了该方案的可行性。最后,通过完成网络化运动控制系统的驱动部分的硬件及运动控制软件的设计,结合之前设计的从站网络控制器及DSP控制器,构建了完整的网络化运动控制系统并将其应用于实际的造波机系统。通过一系列测试,包括电机控制性能测试,造波系统的单轴跟踪和多轴同步性能测试,以及实际造波能力的测试,证明所设计的造波机网络化运动控制系统性能优良,运行可靠。更多还原

【Abstract】 As the marine resources are exploited in depth, people desire more knowledge about ocean. The wave maker system, which is important equipment for simulating wave, is playing more important role in marine research. Wave maker system is large-scale automation equipment which is developed from multi-disciplinary integration such as wave theory, automatic control and mechanical engineering. Base on the fundamental principle of automatic control, motion control system is very significant in system building and could determine the performance of wave maker system. Moreover, The emergence of the networked motion control system brings new opportunities and technical challenges to the developing of wave maker system. Under this background, this dissertation further studies networked motion control system as the purpose for devising a high performance networked wave maker system. The specific content includes:First of all, motion control model of a wave maker system is established according to the actual physical system and the characteristics of system load and moment of inertia is analyzed. Combined with vector control and space vector pulse width modulation technique, an adaptive PID control algorithm is proposed based on wavelet neural network to overcome the difficulties encountered when applying traditional PID to a nonlinear, strong coupling, and multivariable motion control system. This control method well utilizes the good function approximating ability and adaptive capacity of wavelet neural network, and provides a new idea to solve the control problems existing in complex nonlinear, uncertain system. In order to improve the tracking accuracy of the position servo system, an adaptive backstepping control strategy based on wavelet neural network is proposed considering both current and position tracking performance so as to ensure the global asymptotic stability of the system variables and obtain good dynamic and static performance and robustness. On this foundation, sliding mode control is adopted in design to further improve the disturbance rejecting capability of the system. Meanwhile, new wavelet neural network architecture is applied to accelerate the system convergence. Therefore a new adaptive sliding mode backstepping control strategy is obtained, and the simulation results prove that the strategy has superior tracking performance and robustness.Then, networked wave maker motion control system is analyzed and modeled including single-axis and multi-axis model. For the problems brought to the system by network delay in model analysis, wavelet neural network is proposed to predict the network delay, which is able to output the predicted next sample latency value according to past delay sequence. This neural network is updated in real-time via training algorithm so as to ensure the accuracy of the predicted output. Through the experimental analysis for the actual Ethernet delay data, it shows that this prediction model can effectively estimate network delay, so it can be used in actual application. Delay effect in the motion control system is analyzed and Smith predictor is considered to remove the delay out of closed-loop system to ensure the stability of system. In order to overcome the shortcoming that Smith predictor is sensitive to model parameters, an adaptive predictive control algorithm with disturbance compensation is proposed. This algorithm utilizes wavelet neural network to adaptively adjust the reference model output to approach the actual system and feedforward the system disturbance. Simulation results show the proposed algorithm can effectively suppress the disturbances, automatically compensate for model uncertainty and has good dynamic and static performance. To satisfy the multi-axis synchronous control requirements of wave maker motion control system, an adjacent coupling sliding mode synchronous control strategy is proposed, suppressing the uncertainty of the system by sliding mode control in controller designing, to obtain good position servo synchronization performance by eliminate the tracking error and synchronization error together.Further more, a networked motion control system based on the EtherCAT standard industrial Ethernet protocol is constructed and the hardware and software design of the network controller and DSP controller is accomplished. The communication between the master and the DSP controller is realized by using real-time industrial control software in the master and network controller. DSP controller communicates with network controller and realizes the motion control algorithm. When the control commands from master are transmitted via the network to the controller, the controller makes use of the fuzzy PID control algorithm programmed inside to generate the drive signal, which is converted via differential circuit and transmitted to the driver for controlling the movement of the motor. Experimental results demonstrate the performance of this networked motion control system is good and prove the feasibility of this design.At last, on the basis of network controller and controller designed on previous text, a networked motion control system is constructed by designing the hardware of drive part and motion control software, and it is applied in actual wave maker system. A few of experiments is carried out including motor control performance test, single-axis tracking and multi-axis synchronization performance test and actual wave making capability test in experimental flume. The results demonstrate the proposed servo motion control system has excellent performance and reliable operation.更多还原