Fault Tolerant Algorithms for In-orbit Manufacture

在轨制造有很大的动力，可以帮助实现雄心勃勃的大规模月球及更远的任务，并用于电信和地球观测等天基服务。然而，在制造过程中的故障检测和这些结构直接在轨道上的自主组装方面，该领域存在几个问题。研究在空间环境中制造物体的故障检测和恢复方法是本项目的主要重点。将努力确定使用物理传感器，机器视觉和其他输入是否可以正确检测3D打印对象的不同故障模式，例如分层和表面/结构缺陷。这将需要在整个打印过程中模拟和比较打印机的动态，以及打印对象本身在任何阶段的视觉比较。这个最初的工作主体将通知项目的其余部分。预计将需要机器视觉来检测与打印机无关的故障。在这个领域只有有限的研究。预计打印机上的物理传感器和机器视觉算法的结合将能够实时检测打印对象中的故障和缺陷。一旦开发出一种鲁棒的故障检测方法，就可以研究评估和恢复打印对象的方法；如通过机械臂去除杂散材料，或在失效区域添加材料。同样，这一领域的可用工作有限，并且可以设想多种解决方案。最初的努力将基于故障的严重程度。例如，小的缺陷，如翘曲，可以通过添加单层进行补偿来实时恢复。决定错误严重程度和如何恢复错误的算法将是这项工作的重要成果。恢复一个物体需要对模拟技术进行大量研究，以确定哪种方法（如果有的话）将允许打印继续。优化算法可以用来确定恢复的最佳方法，而对于更严重的缺陷，使用机器人操纵器可以去除物体的受影响区域进行重印。由于目标应用旨在支持在轨制造，极端空间环境将对设计解决方案提供限制。考虑到任务的性质，手动干预是不切实际的，发射原料的成本是如此之高，打印成功率尽可能高是至关重要的。在大规模结构制造的情况下更是如此。

There is a large drive for in-orbit manufacturing to help realise ambitious large-scale missions to the moon and beyond, and for use in space-based services such as telecommunications and Earth observation. However, there are several issues within the field related to the detection of failures within the manufacturing process and the autonomous assembly of these structures directly on orbit.Investigating methods for the failure detection and recovery of objects in manufacture within a space environment is the prime focus for this project. Effort will be made to determine if the use of physical sensors, machine vision, and other inputs can correctly detect the different failure modes of a 3D printed object, such as delamination and surface/structural defects. This would require simulation and comparison of both the dynamics of the printer throughout the printing process, and the printed object itself at any stage for visual comparison. This initial body of work will inform the rest of the project. It is expected that machine vision will be needed to detect non-printer related faults. There is only limited research available within this area. It is envisioned that a combination of physical sensors on the printer and the machine vision algorithms will be able to detect faults and defects within the printed objects in real-time.Once a robust method for detecting faults has been developed, methods to evaluate and recover the printing object can be investigated; such as the removal of stray material via robotic arm, or addition of material in the failure region. Again, there is limited work available in this area and there are multiple solutions that can be envisioned. Initial efforts will be based on the severity of the fault. For example, small imperfections, such as warping, may be recoverable in real-time with the addition of single layers to compensate. Algorithms for deciding on the severity of the error and how the error can be recovered will be a significant output of this work.Recovery of an object requires significant research into simulation techniques to determine which, if any, methods will allow a print to continue. Optimisation algorithms can be used to establish what the optimal method of recovery should be, while, for more serious defects, the use of a robotic manipulators could remove affected regions of an object for reprinting.Since target application is intended to support in-orbit manufacturing, the extreme space environment will provide constraints upon the design solutions. It is vital that the print success rate is as high as possible given the nature of the mission where manual intervention is impractical and the cost to launch feedstock is so great. This is even more so in the case of large-scale structural manufacture.