Neural Network Modeling of Turbofan Parameters

Abstract

This research involves the development and validation of neural network models for several engine parameters for a turbofan engine. The investigation encompasses data collection, data translation, development of the neural network models, and testing. The data used to train and validate the neural network models was acquired from state of the art models as well as flight tests. During this study, different neural network architectures and training algorithms are exercised and evaluated for a turbofan engine operating at steady-state conditions. In addition, studies are performed to optimize the neural network architecture and resolution of data used for training. The resulting models are thoroughly validated using data for approximately sixty thousand flight conditions. The neural network models trained and tested with data acquired from state of the art models are capable of predicting their respective parameters with a maximum of 5.3 percent error. A neural network model created with a small set of flight test data and validated with a slightly larger set of data resulted in a maximum error of 4.6 percent.