Pattern Recognition for Time Series Forecasting: A Case Study of a Helicopter’s Flight Dynamics

Abstract

This paper presents a method for time series forecasting based on pattern recognition. As the system receives samples of time series, each of them representing one variable from the set of variables that describe the behavior of an application model, these samples are evaluated using a PCA algorithm, where each sample is represented by a feature vector. Different feature vectors (each of them representing a different sample of a particular case) are compared for pattern recognition. Once this sequence of steps is well performed, it’s possible to estimate time series for different states between those represented by the previously analyzed samples. As an example for application of this method, a case study is presented for some variables under specific flight conditions. The chosen application for this case study, helicopter flight dynamics is a relevant study, for it can be used, for example, to provide precise data to a flight simulator, which implies in an important issue for pilot training, and subsequently, this type of application may help reducing the probability of pilot's faults in real flight missions. To demonstrate the applicability of the method, this paper shows results obtained when the system generated forecasts for flight dynamics variables in a specific scenario of initial conditions and while the helicopter performed a maneuver of response to collective command. Finally, some considerations are made about the work shown in this paper as the results, discussions and conclusions are presented.