面向大型储罐立体巡检的四旋翼无人机自主控制方法研究

Large-scale tank surface quality inspection is one of the important tasks in the storage and transportation industry. At present, this inspection mainly relies on human inspection, which is not only inefficient but existing a large number of missing detection. This increases great security risks to the tank operation. A quadrotor unmanned aerial vehicles (UAV) based approach is proposed in this project to take place of human’s inspection. However, UAV auto-inspection encounters a huge challenge mainly caused by the near-flight tank surface reflection turbulence exacerbating the nonlinear and uncertainty of the UAV control system. Firstly, this project will focus on the three-dimensional coverage path planning algorithm, which satisfies various real constraints, is designed considering the characteristics of the tank structure, and the reference path of the UAV inspection flight is obtained. Then, a Gaussian process learning based nonlinear model predictive control method is proposed to achieve autonomous flight tracking control for UAV under the nonlinear turbulence disturbance. This method adopts the Dryden wind field theory to calculate the nominal model of the UAV, and uses the Gaussian process to learn a statistical model which can be updated online to compensate the imprecise nominal model in real time. Safety metric will be integrated into learning algorithm and the safe domain can be calculated online to ensure system safety and improve tracking performance. In summary, this project addresses key problems in auto-inspection of large tanks. It is theoretically challenging, and also of great importance for industry practice.

大型储罐表面质量检测是储运行业的重要工作之一，目前主要依赖人工抽检，不仅效率低下而且存在大量缺检漏检现象，给储罐运行带来了极大安全隐患。本课题提出采用四旋翼无人机取代人工对大型储罐进行自主巡检。由于近距离飞行时罐壁表面反射紊流加剧了无人机控制系统的非线性和不确定性，因此无人机的自主巡检研究具有极大挑战性。本课题首先结合储罐结构特点设计满足各类约束的三维覆盖路径规划算法，获得无人机巡检飞行参考路径，然后针对无人机近距离巡检时不确定性紊流扰动问题，采用基于高斯过程学习的非线性模型预测控制方法实现无人机的自主飞行跟踪控制。该方法引入Dryden风场理论计算无人机名义模型，使用高斯过程回归学习一个可以在线更新的统计模型，以实时补偿不精确的名义模型；将控制系统安全性融入学习过程在线计算安全域保证系统安全性，提高跟踪性能。该项目针对大型储罐自主巡检的关键问题开展研究，具有重要的理论意义和应用价值。

储罐表面质量对储运工作的安全运行具有举足轻重的作用。随着使用年限的增加，储罐外表面可能出现防腐层的开裂、针孔、脱皮、流挂等缺陷。储运行业通常按照管理制度对储罐定期巡检。现有的检测方法全部依赖于人工携带手持仪进行抽检，存在劳动强度大、人为因素多、储罐高处漏检缺检普遍等问题。本课题采用四旋翼无人机取代人工对大型储罐进行自主巡检，主要研究两个方面的内容：一是巡检无人机的三维路径规划问题；二是靠近储罐外表面巡检时的无人机自主控制问题。前者的困难在于立体巡检的三维路径要考虑避障、全覆盖等问题。后者的难点在于近墙面飞行的无人机受到反向大气紊流的干扰，如何实现无人机的位置、姿态最优跟踪控制。.项目执行期内获得了多项成果，大致分为三个方面：一是提出一种立体巡检路径规划方法。二是针对无人机在巡检过程中易受风扰影响的问题，提出了一系列的参数辨识和在线学习方法，进而设计了相应的控制器，使四旋翼无人机的飞行控制性能获得了提升。三是搭建了储罐巡检实验平台，仿真实验验证了上述方法的可行性和有效性。上述研究成果为降低当前储罐巡检劳动强度、提高检测效率、提升行业智能化水平奠定了基础。