Optimization methodologies study for the development of prognostic artificial neural network

In this work, I discuss the implementation and optimization of an Artificial Neural Network based on the analysis of the back-EMF coefficient capable of making ElectroMechanical Actuator (EMA) prognostics. Aircraft manufacturers are increasingly focusing on the use of electromechanical actuators as they have numerous advantages in terms of weight and compactness respect to the technologies adopted at this time. However, they are early technology and for this reason, engineers are still studying the failure modes that characterize this component. The objective of this thesis is to study a methodology for the recognition of faults within the system. To solve the problem my supervisors have thought of implementing a logic, based on artificial intelligence, particularly on artificial neural networks, which allows to estimate the remaining useful life of the system components starting from a training dataset. The neural network learns autonomously the relationships that link the quantities given as input with those in output. However, during learning, the creators need to set the value of the hyperparameters. My job is to show how these values influence learning and how it is possible to optimize the network to make it more performing in terms of computational cost and complexity, so that the variation of hyperparameters improves supervised learning. The future of aviation is certainly based on the "more electric" philosophy. Electricity is the only indispensable energy source for an aircraft. Nowadays, the remote hypothesis of "full electric aircraft" is still under study and yet there are several queries to be clarified. The results are very satisfactory considering the small number of examples present in the available dataset. In the future, I think that we can build a neural network having datasets with a greater number of examples and deeper even though this, as you can read in this/my thesis, does not always turn out to be an advantage. For this reason, optimizing the work is important.