Machine learning based process monitoring and characterisation of automated composites
RIS ID
125577
Abstract
There has been a huge uptake by industry groups to adapt automated fibre placement (AFP) based manufacturing due to it's high level of productivity, accuracy and reliability. The AFP technology merges through several manufacturing stages like cutting, curing and consolidation. The high level of productivity, accuracy and reliability in automated fibre placement (AFP) have opened new markets and applications for high value laminated composite structures. However, from a system engineering perspective, manufacturing of composites using AFP is a complex, high-dimensional nonlinear multivariable process that involves large number of variables and parameters. The quality and integrity of the structure is critically dependent on the choice of these parameters, which are typically extracted by conducting several lab-based experiments with varied processing parameters. Appropriate selection of these parameters would provide optimal result. Artificial neural network (ANN), a Machine Learning technique has been gaining popularity in various engineering applications including prediction, control, fault diagnosis etc. In this study, a multi-layer perceptron based ANN has been trained to accurately represent the complex relationship between various processing parameters in AFP that would give optimised outcome. The ANN model will subsequently be used to obtain the optimised parameters that can be integrated in AFP based manufacturing of laminated composite structures.
Publication Details
E. Oromiehie, B. Gangadhara. Prusty, G. Rajan, C. Wanigasekara & A. Swain, "Machine learning based process monitoring and characterisation of automated composites," in International SAMPE Technical Conference, 2017, pp. 398-410.