受物理安全感知约束的群系统协同编队

Operational safety and coordinated formation are two main factors affecting the autonomous operation of swarm systems such as the unmanned aerial vehicle swarm and the naval ship swarm et al. Within complex operating environments, physical equipments with high physical safety is necessary for ensuring the coordinated formation of physical swarms. However, the seriously decreasing of physical safety of equipments strongly affects the possibility of formation in swarms. From the remaining useful lifetime representing the physical safety of systems, the project will study the coordinated formation with the constraint of physical safety sensing. To begin with, within the framework of dynamical systems with hidden degradation process, the diagnosis of intermittent faults in physical equipments is studied in a strictly theoretical way. Then, provided that the degradation process is affected by the number of times intermittent faults occur, the remaining useful lifetime can be one main index representing the physical safety of physical systems. At last, constrained by anticipated performance of the probability density function on remaining useful lifetime, the problem of finite-time coordinated formation in swarms is considered in a stochastic sense. We also study the effects of the costs of control and maintenance on coordinated formation. In this project, we try to obtain theoretical results driven by realistic physical properties such as remaining useful lifetime, and verify the effectiveness using the UAV rotorcraft swarm. The main research can provide a theoretical foundation for the autonomous operation of physical swarm systems.

对于无人机群、舰艇群等实际物理群系统，物理安全与协同编队是影响自主运行的重要因素。在复杂运行环境下，成员良好的物理安全是实现群系统协同编队的前提，但物理安全的下降有可能严重影响协同编队的实现。本项目从设备剩余使用寿命这一表征成员物理安全的基本属性出发，研究受物理安全感知约束的群系统协同编队。首先，在具有隐含退化过程的动态系统的框架下，研究成员设备间歇故障的诊断问题，提出严格理论意义下的间歇故障诊断方法；进一步，在间歇故障发生次数影响设备性能退化的前提下，研究成员设备剩余使用寿命的预测问题，并将剩余使用寿命作为评价成员物理安全的主要指标；最后，在设备剩余使用寿命分布函数的约束下，研究随机意义下有限时间内的协同编队问题，并综合考虑维护费用、控制成本对协同编队的影响。本项目致力于取得实际物理属性驱动的理论研究成果，并在旋翼无人机群实验平台上进行验证，从而为实际物理群系统的自主运行提供理论基础。

项目组主要围绕设备故障诊断、设备剩余使用寿命预测以及群系统协同编队进行了系统性的研究，基本上完成了预期的研究目标。主要取得了如下两项重要成果。..第一项重要成果：基于变元的微小故障检测方法。..由于微小故障具有幅值小、持续时间较短等现象，PCA方法很容易将微小故障看作噪声。项目组直接对测量矩阵进行正交变换，得到了完全正交的变元，能够设计对微小故障异常敏感的指标。借助变元，能够得到了一系列创新性的成果。（1）将变元与时滞技术相结合，提出了动态递推变元方法，能够对田纳西伊斯曼过程（TEP）故障15的检测率超过95％，远优于现有PCA方法以及变体。（2）基于递推变元统计分析，提出了传感器早期故障的分离方法。与重构贡献图等方法相比，具有更优越的分离性能。（3）基于变元的增量，递推变元能够解决实际高炉炼铁过程中的悬料异常检测问题，极大减少了过程的时变特性的影响常。上述成果发表于Automatica、IEEE Transactions on Industrial Electronics、Journal of Process Control以及Control Engineering Practice上。..第二项重要工作：基于分形布朗运动的剩余使用寿命预测理论..从实际大型高炉的退化特点出发，凝练出一类全新的性能退化模型，在国际上创造性建立了基于分形布朗运动的剩余寿命预测理论，克服了传统的基于布朗运动的预测理论的严重缺陷，开辟了剩余寿命预测的一个新方向。绝大多数现有文献仅仅采用无记忆效应的马尔科夫过程来建模退化过程。为了描述实际存在的记忆效应，退化模型中的扩散项由分数布朗运动（FBM）来表示。FBM实际上是一种特殊的非马尔科夫过程，且具备长程相关性。上述成果已经在IEEE Transactions on Reliability期刊上发表和录用论文3篇。

内容获取失败，请点击重试