A numerical tool for smart in-situ sensing of defect features in large scale infusions
Topic(s) : Special Sessions
Co-authors :
Jack M DAVIES (UNITED KINGDOM), Dmitry S. Ivanov (UNITED KINGDOM), Peter GIDDINGS , Janice M. DULIEU-BARTON (UNITED KINGDOM), Andrew JENKINSAbstract :
Recently, wind turbine blades have reached record breaking lengths (and build rates) with the innovative use of composite materials. However, such large composite parts increase the manufacturing complexity and risk. The vacuum infusion process used to manufacture large scale composite parts like wind blades is susceptible to air entrapment, which often results in resin deficient regions or ‘dry spots’ that require costly and wasteful repairs.
The prediction of air entrapment in non-transparent moulds at blade scales remains a significant challenge. The fidelity of computer simulations is governed by the accuracy of the material characterisation and the modelling assumptions. Permeability can vary due to design features (e.g. ply drops), process variations such as decompaction, ply/core misplacement, and intrinsic material variability (e.g. yarn loss). As a result, drastically different flow patterns can be experienced during manufacture often leading to unexpected dry spots, which compromise structural performance and require remedial work before the part goes into service. Therefore, a means of producing more consistent resin infusion parts is proposed, which is achieved by improved monitoring, simulation and control of the process. Simulations that better reflect the physical process allow more confident control decisions for real time defect mitigation.
A numerical approach is developed that determines the size and position of potential defects. The simulation is based on the effect of permeability variations on flow front progression and pressure evolution, derived from Darcy’s law and mass conservation in 1D. The simulation procedure optimises the parameters in the governing equations using a nonlinear least squares regression so that the pressure history in the model matches the pressures obtained from sensors in the mould tool.
Figure 1 shows that the simulation is able to build an accurate description of high or low permeability defects present in an infusion, providing that there is an adequate distribution of sensors. The effect of sensor placement on the accuracy of defect characterisation is shown in Figure 2. Sensors are best placed ahead of and within the sensing range of the simulated defects’ position. The effective range of the sensors is a trade-off between the sensor distance from the inlet and its measurement resolution.
In summary, the research demonstrates the feasibility of using a simple tool to optimise sensor position and distribution to detect flow features in real time. A practical challenge is the high density of sensors that is required to detect the small features that may cause air entrapment and require defect mitigating actions. In this work we discuss an approach to balance computational efficiency of feature detection against more elaborate true-to-life models, and the potential for finding a pragmatic and efficient strategy for defect mitigation at large structural scales.
The prediction of air entrapment in non-transparent moulds at blade scales remains a significant challenge. The fidelity of computer simulations is governed by the accuracy of the material characterisation and the modelling assumptions. Permeability can vary due to design features (e.g. ply drops), process variations such as decompaction, ply/core misplacement, and intrinsic material variability (e.g. yarn loss). As a result, drastically different flow patterns can be experienced during manufacture often leading to unexpected dry spots, which compromise structural performance and require remedial work before the part goes into service. Therefore, a means of producing more consistent resin infusion parts is proposed, which is achieved by improved monitoring, simulation and control of the process. Simulations that better reflect the physical process allow more confident control decisions for real time defect mitigation.
A numerical approach is developed that determines the size and position of potential defects. The simulation is based on the effect of permeability variations on flow front progression and pressure evolution, derived from Darcy’s law and mass conservation in 1D. The simulation procedure optimises the parameters in the governing equations using a nonlinear least squares regression so that the pressure history in the model matches the pressures obtained from sensors in the mould tool.
Figure 1 shows that the simulation is able to build an accurate description of high or low permeability defects present in an infusion, providing that there is an adequate distribution of sensors. The effect of sensor placement on the accuracy of defect characterisation is shown in Figure 2. Sensors are best placed ahead of and within the sensing range of the simulated defects’ position. The effective range of the sensors is a trade-off between the sensor distance from the inlet and its measurement resolution.
In summary, the research demonstrates the feasibility of using a simple tool to optimise sensor position and distribution to detect flow features in real time. A practical challenge is the high density of sensors that is required to detect the small features that may cause air entrapment and require defect mitigating actions. In this work we discuss an approach to balance computational efficiency of feature detection against more elaborate true-to-life models, and the potential for finding a pragmatic and efficient strategy for defect mitigation at large structural scales.