Identification of Helicopter Dynamics Using Recurrent Neural Networks and Flight Data

A comparative study and analysis of different Recurrent Neural Networks (RNN) for the identification of helicopter dynamics using flight data is presented. In RNN based techniques, the model structure need not be defined apriori unlike in the conventional identification methods such as least square estimation, quasi‐linearization and stochastic modeling. In case of a helicopter, defining an apriori model is difficult because of the interplay between various subsystems such as the rotor, fuselage, power plant, transmission, empennage, and tail rotor. For a given input‐output data set, RNN are able to capture the underlying relationship. Three different RNN architectures, namely, Nonlinear Auto Regressive eXogenous input (NARX) model, neural network with internal memory known as Memory Neuron Networks (MNN), and Recurrent MultiLayer Perceptron (RMLP) networks, are used to identify dynamics of the helicopter at various flight conditions. Flight data of a four bladed helicopter with soft‐inplane hingeless rotor is used for the simulation studies. Each RNN model is used to approximate the flight data at different flight conditions. Results of identification of the coupled and uncoupled (longitudinal and lateral separately) dynamics are compared. The performances of different RNN architectures are studied. Based on the results, the practical utility, advantages and limitations of the three models are critically appraised and it is found that the NARX model is most suitable for the identification of helicopter dynamics.