Advanced Metrology Hub for Sustainable Manufacturing

The second stage will create new, rapidly deployable mechatronic control models fusing multiphysics-based, predictive ML and large-language models to achieve evolutionary self-learning and polymorphic applications with inputs from all levels of embedded and post-process metrology, key challenge is how to control output quality, ML trust deficit, and lack of physics consistency. The raw data collections from activities conducted the first stage (as well as collected data from other Hub projects) will provide inputs for these models, allowing a fusion of modelled physics and physical data. Methods of stress-testing control models through adversarial and penetration testing techniques will be devised. 

The final part of this project will combine constrained generative design and topological optimisations to exploit the findings from the first two stages, resulting in a control architecture capable of assured self-calibration. Flexibility, reconfigurability, evolvability and robustness to different manufacturing requirements will be validated and complex time-varying multivariable interference will be created, for priority application within the Hub’s other sustainable manufacturing ‘strands’.