【作者】 王磊；

【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， 2012， 博士

【摘要】 作为自动控制领域的一个重要分支,运动控制技术已被广泛应用于工业生产、设备制造等多个领域。随着控制任务复杂化、控制设备多样化,传统的点对点的控制方式已不能满足其大规模、分布式、多轴协同的控制要求。将网络引入该领域构成分布式网络运动控制系统,不但可以解决这些难题,而且能够简化系统的软硬件设计,降低成本。考虑到网络运动控制技术在大型海工实验设备—海洋造波机中的应用,本文拟研究网络延迟、丢包等对系统性能的影响,通过系统模型的建立和控制技术的研究解决吸收式造波机网络运动控制系统高速实时控制、高精度同步等关键技术难题,具体从以下几个方面做深入的研究：首先,从造波机网络运动控制的特点切入,构建层级式结构的海洋造波机网络运动控制系统,基于该结构通过对一个造波机网络运动控制系统子单元的分析和建模,给出在一定约束条件下整个系统全闭环数学模型。对该模型在四种应用触发模式框架下进行分析,明确建立的系统的触发模式。考虑网络运动控制系统现场应用的特点,进一步放宽约束条件,在系统中引入随机噪声干扰信号,并注意到在层级式结构下,数据包出现乱序的可能性,建立造波机网络运动控制系统现实的统一的数学模型。分析该模型,确定系统的稳定条件,并利用求解线性矩阵不等式的方法证明给出条件的合理性。其次,为了满足实验需要,确保在有界时延内检测到现场造的波浪的参数并传送给上位机,用以修正目标波谱数据,产生下一采样周期的控制命令,控制造波板做相应运动,实现主动吸收功能。考虑到在提出的层级式网络运动控制系统下,通信协议能够通过标准的以太网数据帧携带时间戳,因此可以在上位机记录端到端的网络通信延迟。基于此建立采用神经网络的观测预测器模型,预测系统未来延迟,从而确定系统下一采样时刻控制指令的输出。具体实现思路是将记录到的第K次采样之前的延迟时间序列值作为神经网络的输入,通过前馈神经网络进行训练,预先得到第K+1次采样的系统时间延迟值,根据此值给出系统控制输入的预估值。该方法实现的关键有两点：一是所采用的神经网络模型拓扑结构的选择,二是预测观测器模型的建立。此外,对提出的模型进行仿真分析,以验证基于该模型的系统在存在一定的延迟和丢包的情况下的稳定性及模型的有效性。再次,对在不同网络延迟类型下的系统模型的PID参数进行优化设计,给出符合评价函数标准的P、I、D几个参数在不同采样时间和系统时间常数下的变化规律。在分析不同类型延迟影响的基础上,为了实现系统的PID参数在线整定,提出两种网络PID智能整定方法：基于二进制编码的遗传算法PID整定方法和基于模糊控制的PID整定方法。针对相关整定方法建立系统仿真模型,在给定的网络延迟情况下,对比智能PID整定和常规方法整定的效果,分析延迟的引入对系统响应的影响。最后,采用时空图方法,分析基于三种主流工业以太网协议的系统时延构成,通过对比包含不同数量从站的系统通信延迟的变化,从理论的角度重点评估EtherCAT协议的延迟性能。同时设计基于EtherCAT协议的时钟同步功能实现方案,研究新的时钟漂移动态补偿算法。在此基础上,提出将EtherCAT协议用于海洋造波机网络运动系统的现场层控制,更高层直接连接通用以太网的层级式设计方案,在方案基础上搭建系统实验平台。为了保证实验平台的下位主站在多任务操作系统Windows下实时控制功能的实现,探讨WinXP操作系统之上的INtime实时核扩展,以及在该框架内主站协议栈功能的实现；设计采用高性能DSP芯片加上网络专用接口芯片以及控制接口电路的完整从站硬件平台,开发从站协议栈及控制应用程序,实现协议栈在DSP上的移植。系统设计拟实现以下目标：(1)构建多功能高速大带宽数据采集系统,使子系统主站能够对造波板前的波况信息在低于1毫秒的周期时间内进行实时同步采集；(2)搭建鲁棒性较好的闭环造波机网络通信控制系统,实现实时多板同步控制；(3)以网络为数据传输媒介完成对各个造波板的同步控制,并在多任务操作系统下实现造波机系统的实时控制。为验证建立的实验平台的有效性和可靠性,需要对建立的系统进行两方面的性能测试评估。即：一方面,通过多次造波实验,测试系统造波的稳定性、重复性和精确性,另一方面,对下位系统进行周期性和同步性的测试。更多还原

【Abstract】 As one of the important branch of automatic control field, motion control technology has been widely used in modern industrial production, design, manufacture, and so on. However, with the complexity of control tasks and diversification of control devices, the traditional point-to-point control mode cannot cope with a large, distributed, and more collaborative axis control requirements. Introducing communication network into motion control field to build distributed motion control system can not only solve the above mentioned problems but also simplify the design of the software and hardware system and reduce the cost of system. Taking into account the application of networked motion control technology for large-scale experimental equipment named Wavemaker in ocean engineering field, the paper will try to solve the key technical problems which the absorption Wavemaker is controlled with high accuracy and high real-time performance. The model is established aiming at the delay time and data packets drop over network. And a series of intelligent control technologies are researched. The paper is outlined as follows:First of all, in view of the characteristics of Wavemaker, the hierarchical system structure of networked control Wavemaker is proposed. The system mathematical model is given based on analyzing one of control units of the subsystem. And the all closed loop system model is proposed under certain condition. The system trigger mode should be clearly defined in the framework of four application trigger modes. Considering the characteristics of networked motion control systems, the paper presents a unified mathematical model of networked motion control system for Wavemaker. In the model, the constraint condition is relaxed. The system with interference signal and disordered data packets is analyzed. The system stability conditions are proposed for the above mentioned mathematical model. The conditions are proved by the method of linear matrix inequalities.Second, to revise the target spectrum and cope with the experimental requirements, the wave parameters must be timely tested and sent to the upper computer under the bounded delay time. Taking advantage of the characteristic which the timestamp can be carried by the data packets in the networked motion control system with hierarchical structure, the observer prediction model based neural network is presented according to the recorded end-to-end communication delay in the upper computer. The model is used to predict the next communication delay time to determine the required control instructions. The specific method is to record the past delay time sequence values before the data sampling time K. The values are entered into the neural network and trained by the neural network model. After the training process is finished, the new control instructions are generated by the predicted delay time values before the data sampling time K+l. The proposed method includes two key aspects:the neural network topology selection for the predicted model. Moreover, taking into communication delay and data packets drop, the system based the model should be stable and the effectiveness of model must be verified through the simulation method.Third, the PID(Proportional-Integral-Derivative) parameters of networked motion control system should be optimized under different types of delay time. At the same time, the variation of parameters by cost standard function is given according to different sample time and system time constant. To achieve the online PID parameters of the subsystem automatic adjustment, the intelligent networked PID adjustment methods are proposed based on the analysis of the different types of delay time. Two methods are put forward named PID control method based on binary coded genetic algorithm and PID control method based on fuzzy theory, respectively. The system simulation model is established to evaluate the difference between the conventional adjustment method and the intelligent PID regulation method under Pre-set delay time.Finally, the composition of Industrial Ethernet delay time is analyzed by space-time diagram. We compare the delay time difference of the system including different number of slave stations using three mainstream Industrial Ethernet protocols, respectively. And we focus on verifying the EtherCAT protocol performance from the perspective of the protocol theory. And clock synchronization scheme based on EtherCAT protocol is put forward and the new clock dynamic drift compensation algorithm is set forth. Next, the design scheme of networked motion control system for Wavemaker is proposed. In the scheme, the EtherCAT protocol is proposed to construct the field devices control network and these devices communicate with the upper computer through standard Ethernet. The experimental platform is set up based on above scheme. The whole system structure consists of two parts named master station and slave station. To ensure the realization of the real-time performance, the function of EtherCAT master station protocol stack should be achieved on the Windows system platform which is extended the INtime real time core. The paper gives the hardware design diagrams of EtherCAT slave station experiment platform based on DSP controller and EtherCAT network interface chip. In addition, the software achieves the function of EtherCAT slave station protocol stack and the flow chart of each program module is presented. The system should be able to achieve the following goals:(1) The high-speed and high bandwidth data acquisition system should be constructed, and the master station of subsystem can finish real-time synchronization of data acquisition less than1millisecond.(2) To control synchronously the wave making boards, the closed-loop Wavemaker networked communication control system with better robustness should be set up. (3) The system must be able to control synchronously all of wave making boards through the communication network. And the real-time control function should be achieved under multi-tasking operating system.To verify the effectiveness and reliability of the designed experimental platform, the system performance must be evaluated. We test the stability, repeatability and accuracy of the designed system through making repeatedly wave in the experimental pool. On the other hand, the cycle and synchronization performance indexes of the subsystem should be tested.更多还原