Targeting cost-effective and efficient infusion control with real-time sensing and modelling

Abstract Commercial trends show continuous increases in the production rates and scale of wind turbine blades to meet the rising demand for cost-effective renewable energy. Manufacturing defects undermine such efforts by reducing turbine performance and causing delays, downtime and potentially millions of euros in repairs costs per turbine. Dry spot defects arise due to variations in the resin infusion process, demanding more adaptive manufacturing techniques. However, the real-time information required on the resin flow front is often unavailable or limited in industrial-scale parts and tooling setups. The number of sensors required in a blade mould (hundreds of metres in length) is impractical and remains a critical challenge. The aim is to provide a solution by maximising the information obtained from the fewest possible sensors. It is shown that just one pressure sensor, together with a smart physics-based computer model, can enable online estimations of the spatially varying material properties, even when the variations occur metres downstream from the sensor. Then, the position of resin flow front can be predicted with high accuracy as shown by two of the three virtual infusion experiments, where the flow front was predominantly orthogonal to the infusion direction. The devised approach addresses the challenges of adopting sensors in large-scale serial production processes, allowing informed process control decisions for the defect-free manufacture of wind turbine blades.