【作者】 肖金壮；

【导师】 王洪瑞； 宋维公；

【作者基本信息】 燕山大学 ， 控制理论与控制工程， 2010， 博士

【摘要】 机器人在当今世界得到不断广泛而深入的应用,与此同时机器人系统安全性指标成为了人们在应用过程中关注的重要方面。机器人故障的发生具有很强的随机性,并且随着使用时间的增加,发生故障的频率会呈指数上升,使得对故障的探测、诊断以至容错方面的研究意义明显。针对机器人系统安全性方面的研究虽然已经有了一些的研究成果,但仍存在较多的关键问题需要解决,同时就系统中编码器和执行机构相关复杂故障的探测与诊断过程,仍需要行之有效的方法。另外,在机器人系统中驱动故障的容错方面,智能性和适应性也需要提高。针对这些机器人系统中安全相关的实际问题,本文开展了一下几方面的研究工作:(1)在机器人系统的故障探测方面,针对难以探测的一般机器人系统位置环反馈装置编码器的丢码、漏码故障,利用系统中冗余的电机转速信号,提出了基于主元分析法(PCA)的编码器故障探测方法。通过数据仿真证明了该方法的有效性和方便性;另外,考虑到一般机器人系统中电机工作于速度控制模式,当机器人系统发生扭矩突变故障时,伺服系统不出现报警的问题,提出了将机器人非线性动力学模型作为非线性核函数的思想,利用主元分析法探测位置信号与扭矩信号的相关性,并根据异常的不相关性探测故障的发生。该方法的有效性在X-Y平台与外界的碰撞故障探测过程中得到验证;同时,针对机器人系统中可能发生的机器与机器、机器与人等外界环境间的碰撞故障,提出了基于小波函数故障探测方法,能够反映出不同碰撞的性质,在以X-Y平台为对象的实验过程中验证了方法的有效性;(2)在基于机器学习的故障诊断方面,针对机器人系统中位置/转角与扭矩间的非线性关系,通过向机器人故障诊断算法中引入非线性支持向量机,解决对驱动失灵故障、连接类故障以及与外界碰撞故障的多模式故障诊断问题,并提出基于此思想的机器人系统一般故障诊断框架,以实际过程验证了有效性;(3)在机器人驱动系统故障的容错控制方面,针对输出的饱和故障,为了在有限的驱动能力下,保证系统稳定,同时兼顾系统的快速性,文中给出了基于模糊规则的主动故障容错方法,并通过Lyapunov理论证明了容错控制系统的全局稳定性,仿真试验证明了方法的有效性。上述故障探测、诊断及容错方法都考虑到了应用过程中的实用性和有效性,并以通用X-Y运动平台为对象进行了实验验证,个别方法通过Matlab软件进行仿真验证,为更好地解决机器人系统运行中的安全性相关问题提供了几种新的思路和方法。更多还原

【Abstract】 Robots are applied in the world abroad and deeply. In this background, the index of robotic safty has becoming the important aspect during human use them. Faults in robotic system have very high randomicity. With the machines become dated, the fault occuring frequency rise exponentially. The importance of fault detection, diagnosis, and tolerant control is evident. Although there are some meaningful research results in the aspect of the safty in robotic system, many pivotal problems need effective solving methods. Especially some complex problems ask for available techniques, for example, the faults in encoder and actuator. In another aspect, the fault tolerance on systemic actuators demands higher intellegence and adaptability. Focusing on these practical problems about the safty of robotic system, the research results can be summarized as follows:(1) On the aspect of fault detection in robotic system, firstly, as the very important feedback equipment in postion loop, encoder could lose codes or pause codes. Benifiting from the redundant signal of motor velocity in normal robotic system, a PCA-based mehod was developped. With numerical simulation, the validity and conveniency was tested.And considering normal robot runs under the velocity mode of driver equipment, when the applied torque turns into unconventionality and works in its nominal area, servo equipment would not give off alarm. A thought of treating the robotic nonlinear dynamic model as nonlinear kerel function was presented, where using PCA to detecting the relativity of the output position signal and torque monitoring signal. According to the noncorrelation of unconventionality, fault can be detected. The efficiency of this method was tested when an X-Y motion platform impacts a roadblock.At the same time, aiming at the contingent collision fault, between machines or robot and outer circumstance, the wavelet-based fault detection method is given, from which property of the collision can be reflected. An experiment was developped to testing the efficiency of this method, where an X-Y platform impacted two different obstacles under same motion velocity. (2) The fault diagnosis method using thought of machine learning, considering the nonlinear property of position/angle and applied torque in robotic system, nonlinear support vecter machine was introduced to diagnosing the faults in robotic system, and the commonly fault diagnosing framework was developed for this system. This method was validated in the X-Y platform where different faults were simulated practically.(3) In order to implement fault tolerant control when robot driver equipment falls into saturation state, a method was advanced using the thought of fuzzy rule adjustment. Under it, the close-loop control system is stable and satisfying the requirement of speedability. The globle stability was proved under the Lyapunov sable theory. The effectiveness of this thought was validated on a two degree of freedom manipulators.All of the above fault detecting, fault diagnosis and fault tolerant control methods were developped under the thought of practicability and effectiveness. They were mainly validated on a classical X-Y motion control platform with open configuration. A few methods validated using Matlab because it was hard to simulating these types of faults. As a conclusion, this work gives some new thoughts and methods for solving the safty-related problems in robotic system.更多还原