The Journey Towards Condition-Based Maintenance: a Framework for the Horizontal Tail Actuator of an Advanced Jet Trainer Aircraft

The emphasis on innovative maintenance strategies is driving the industry towards the analyses of operational and historical data from a new perspective. Condition-Based Maintenance (CBM) and Prognostic and Health Management (PHM) strategies can benefit every stakeholder along assets life cycles in assessing systems health status and limiting unexpected breakdowns, hence enabling optimized planning of maintenance actions, better availability and lower operative costs. This consideration holds particularly true for the aerospace industry as aircraft maintenance costs can account for significant portions of Life Cycle Costs (LCC). Stored in-service operational data, initially collected for other purposes such as Structural Health Monitoring (SHM), could hold significant value when used as comprehensive datasets for the construction of PHM frameworks. On the other hand, unique challenges arise when implementing PHM logics for legacy and operational platforms due to data availability, quality, consistency and the complexity of integrating data from a multitude of sources and formats into a single framework. This paper underscores once again the high-effort high-reward scenario of developing PHM strategies on equipment in operation and highlights the challenges and trade-offs required to deal with an In-Service dataset. A data-driven approach which relies on operational and historical data for an Advanced Jet Trainer (AJT) is reported in this paper. The research project has been tailored to address a specific subsystem: the Horizontal Tail (HT) flight control actuator, which has been thoroughly analyzed starting from design documents and performance values to operational flight and maintenance/logistics data acquired from an actual fleet of as many as 22 aircrafts, reaching more than 25000 flight hours. Following the overview of the available data repository, the customized methodological workflow is shown. After conducting analyses on data quality and sampling, a statistical approach based on cumulative features (CF) has been adopted. The first four statistical moments have been employed as predictors to extract lumped data statistical characteristics. This methodology has been selected to assess whether the available data exhibits predictive significance in regards to the designated subsystem. Finally the main results and next steps of the research project are reported.