Computer Learning in Automated Manufacturing Processes (CLAMPS) demonstrates the integration of predictive machine learning to drive improvements and minimise variability within a composites manufacturing process. A collaborative project between the Centre for Modelling & Simulation (CFMS) and the National Composites Centre (NCC), it highlights the digitalisation and automation steps necessary to ensure consistently high-quality parts, ultimately reducing costs.
The next generation of single aisle aircraft need to meet both performance (e.g. high strength, low weight etc.) and industrial (e.g. high production rate, rapid ramp up rate) requirements. Traditionally, such aircraft have been designed with performance as the key criteria but now there is a requirement for rapid exploration and faster maturity of novel composite manufacturing processes to meet additional industrial requirements. Composite manufacturing is intrinsically a high variability process, requiring careful tuning of process parameters by highly skilled engineers to meet product quality requirements. This has an adverse effect on production rate, ramp up rate, and increases manufacturing costs.
A combination of traditional virtual manufacturing simulation and artificial intelligence technology is a potential solution to rapidly explore and mature novel composite manufacturing processes (e.g. resin infusion into dry fibre preform). The ability to detect and control defect formation was based on the output of over 15,000 virtual manufacturing simulations of a liquid composite resin infusion process. A machine learning algorithm was used to process the data and classify the decisions made during the infusion process to minimise the formation of porosity and dry-spot defects. The flow of resin was monitored using intelligently positioned sensors, enabling a real-time understanding of the flow inside a closed mould. The machine learning model was used to predict and mitigate the formation of defects by selectively opening and closing inlet and outlet valves to influence the flow of resin.
The key benefits of this technology are; insights into manufacturing process parameters leading to accelerated learning curves; automation of manufacturing inspection step leading to reduction in ‘cost of quality’; upfront prediction of product quality from the machine learning model provides pre-emptive power which in turn reduces the ‘cost of non-quality’. It also has the potential to be upscaled to complex composite part geometries for sectors including aerospace, automotive, construction and energy.