Learning tracking controllers under unknown disturbances for the installation of rigid and flexible structures
Learning tracking controllers under unknown disturbances for the installation of rigid and flexible structures
复制标题
在未知干扰下学习跟踪控制器，用于安装刚性和柔性结构
DOI：
10.1007/pl00010628
复制
发表时间：
1999
影响因子：
2.6
通讯作者：
Keisuke Watanabe
中科院分区：
文献类型：
--
作者：
Hideyuki Suzuki;Tao Qi;Keisuke Watanabe
文摘：在水下结构物安装作业中，半潜体等海洋结构物、海流、潮流产生的定常非线性水动力以及结构物之间的水动力相互作用被视为主要扰动。一般来说，事先收集关于这些干扰的信息是困难和非常昂贵的，尽管这种知识对于精确安装是重要和必要的。基于这种情况，我们提出了一种用于近海结构物自动跟踪/停靠作业的学习跟踪控制器。LTC可以在没有任何电流信息的情况下运行，可以应用于柔性结构的安装问题，其灵活性不容忽视。为了验证LTC的有效性，利用未知水流的实验模型进行了水池试验。利用超声波测距系统和推进器进行了两种类型的盆地实验。一个是半潜器的轨迹跟踪，另一个是柔性水下结构的精确对接。在这两种情况下，LTC都表现出了良好的性能，特别是在后一种情况下，成功地建立了两个灵活的模型来跟踪给定的弹道，并以±5 mm的精度对接到对接目标上，该弹道已经被跟踪了6到7次用于学习目的。柔性模型没有激发弹性响应。
Abstract: During the installation operations of underwater structures, offshore structures such as semisubmersibles, the steady nonlinear hydrodynamic force due to the ocean current, the tidal current, and the hydrodynamic interaction between structures are regarded as major disturbances. Generally, it is difficult and very expensive to collect information about these disturbances in advance, although such knowledge would be important and necessary for precise installation. Because of this situation, we have proposed a learning tracking controller (LTC) for automated tracking/docking operations of offshore structures. The LTC can be operated without any information about current, and can be applied to the installation problem of flexible structures whose flexibility cannot be ignored. In order to confirm the effectiveness of the LTC, basin tests were carried out using experimental models with an unknown current. Two types of basin experiment were executed using an ultrasound ranging system and thrusters. One was trajectory tracking of a semisubmersible, and the other was the precise mating of flexible underwater structures. In both cases the LTC showed good performance, and in the latter case in particular, two flexible models were successfully made to track the given trajectory and dock to the docking targets with an accuracy of ±5 mm, after the trajectory had been tracked six to seven times for learning purposes. No elastic responses were excited in the flexible model.