Aircraft Ice Accretion Prediction Based on Neural Networks

Many experimental research efforts in the past two decades have revealed that the complete picture of aircraft ice accretion has many components, resulting in a complex physical structure. Although overwhelmingly complex, the icing phenomenon needs to be understood because of its impact on aircraft performance and safety. This requires a detailed knowledge of ice accretion physics, subsequent flow over the aircraft, and the resulting modified aircraft performance. Experimental and numerical studies to address these issues have their own advantages, disadvantages, and limitations, which further limit the analysis of the icing phenomena. The motivation behind this study is the belief that complex phenomena in nature have an orderly structure on the large scale. Based on this premise, it is thought that icing phenomena also have orderly, albeit nonlinear, behavior that can be modeled by neural networks, which have a proven capability for modeling nonlinear systems. The methodology developed in the present study incorporates the Fourier series expansion of an ice shape following a conformal mapping, which suppresses the effect of airfoil geometry, and then utilizes neural networks to model the Fourier coefficients and the downstream extent of the ice shape. The neural network can be trained to make ice accretion predictions, given a set of data including the flight and atmospheric conditions, along with the Fourier coefficients and the extent of the resulting ice shape. The neural network also provides statistical output of the relative significance of the input parameters in the training. The preliminary results show that the proposed method has reasonable capabilities and has merit for further investment, because it can be coupled with other systems to create advanced computational ice accretion models and ice protection systems.