The aircraft encounters icing weather during flight, especially the wing icing caused by the icing conditions of supercooled large water droplets, which will seriously affect the aerodynamic performance and handling quality of the aircraft, resulting in flight faults or flight accidents. It is an important factor for the flight safety of the aircraft and a difficult problem that must be solved in the development of the aircraft. Although a variety of calculation and simulation programs have been established for icing, it is a solution that engineers have been pursuing to quickly obtain the simulated icing shape, including the icing model under the condition of supercooled large water droplets. This paper attempts to establish a wing icing model in standby mode by using forward multilayer neural network. Based on the coordinate transformation principle, this method converts the obtained wing icing data of the standard airfoil expressed in rectangular coordinates into the data expressed in corresponding polar coordinates. The wing icing model is constructed with flight parameters, meteorological parameters and polar coordinate angles as inputs and polar coordinate modulus as output training. Using the established model, the simulation and prediction results show that the method is fast and accurate, and its calculation accuracy can satisfy the actual requirements.