Development of health features and automatic fault diagnosis methods for primary flight control actuators

As the leading technology employed in primary flight controls, Electro-Hydraulic Servo Actuators (EHSA) have been the focus of several studies in the aviation industry, particularly in the field of fault diagnosis. The established testing procedures at Maintenance, Repair and Overhaul (MRO) companies are manual, time-consuming and do not guarantee an optimal condition-based maintenance. This thesis stems from a research project developed at Lufthansa Technik, with the purpose to automatize and improve the accuracy of the diagnostic procedures. The proposed approach consists in commanding the servoactuators with a modular test signal, which reproduces the tests from the component manuals. In addition, the signal contains a series of original test sequences, created to assess the state of two critical subassemblies: the servovalve and the recentering device. To elaborate the response of degraded units, an automatic feature extraction algorithm was implemented, which applies a statistical technique known as change points detection to achieve a high reliability. The features extracted were evaluated by means of performance metrics that quantify the correlation with the degradations applied on the component. To test the effectiveness of the health features for fault isolation, these were used as the input of different types of machine learning algorithms. The training process was initially performed on simulations obtained from a high-fidelity numerical model, to classify between different types of degradations, and subsequently on historical data from the workshops, to predict the replacement of a subcomponent by the technicians. The final result is an automatic diagnostic system that integrates an adaptable excitation signal, health features closely related to the physics of the components and a classification algorithm capable of identifying different failures. This innovative framework aims to increase the productivity of the workshops and to determine an improvement in the quality of the repaired components, which would lead to longer times between overhauls.