Elastic Manufacturing Analytics: Improving aerospace assembly processes in an evolvable cell through adaptive data analytics and machine learning

可重构单元有望提供新一代制造系统。这些细胞被设计成能够适应不断变化的环境。利用数据做出决策，这些单元将逐步重新配置，以更好地适应生产需求。然而，目前还不清楚当前的数据分析系统如何支持这些复杂的环境。随着资产的变化，我们可能会随着新传感器的添加而获取新类型的数据。将其与遗留数据聚合可能是无效的，因为数据集可能在质量、数量或类型上有所不同。这些数据集用于训练机器学习模型。未经处理的数据集可能会导致不需要的输出或过度拟合。这些输出对于做出适当的重构决策至关重要。本研究的目的是双重的。首先，是确定要获得的相关数据类型。其次，开发利用自适应数据分析和机器学习方法的数据系统，用于复杂环境中的可重构单元。这项工作将为未来制造系统的学习能力提供基础，并随着时间的推移提高重构和弹性。特别是，它将成为国家示范者全能工厂的支柱，这是一个独特的可重新配置的组装设施，旨在成为未来项目的国家试验台。

Reconfigurable cells promise to deliver a new generation of manufacturing systems. These cells are designed to be adaptable in changing environments. Utilising data to make decisions, the cells will incrementally reconfigure to be better suited to production demands. However, it is not understood how these complex environments can be supported with current data analytical systems.As assets change, we may acquire new types of data as new sensors are added. Aggregating this with legacy data may be invalid as the datasets could vary in quality, quantity, or type. These datasets are used to train machine learning models. Unprocessed datasets may cause undesired outputs or overfitting. These outputs are essential to make appropriate decisions of reconfiguration.The goal of this research is twofold. Firstly, it is to identify the relevant type of data to obtain. Secondly, it is to develop data systems that make use of adaptive data analytics and machine learning methods for reconfigurable cells in complex environments.This work will deliver the foundations for future manufacturing systems ability to learn and improve reconfiguration and resilience over time. Particularly, it will become a pillar of the National Demonstrator Omnifactory, a unique reconfigurable assembly facility that aims to serve as a national test bed for future projects.