In predictive maintenance the schedule is planned based on the actual condition of the asset, and its life projection to the future. In order to implement predictive maintenance, the remaining useful life must be computed. An approach to do so, is via data-driven modelling. The reason behind predictive maintenance, is that a suitable maintenance schedule can reduce significantly the maintenance costs and use of material, by increasing the availability and the reliability of a gas turbine.The objective of this work is to create a remaining useful life prediction model for gas turbines based on data-driven approach. Data-driven models rely on existing information analyzed using data-analytics. That be simple statistics or artificial intelligence. The benefit of this method is the exploitation of all the available data acquired by operators from the onboard sensors through the diagnostics system.For this work an open-source database is used. The dataset is composed by virtual sensor measurements, computed through a thermodynamic model for a given turbofan engine. First the necessary data pre-processing is done, taking into account data corrections with respect to operating conditions. Noise reduction is performed via averaging methods such as moving mean average. Then feature selection takes place, based on the correlation matrix and the utility of each sensor for describing the physics of the problem.The selection of algorithms tested is based on literature review. The selection of their hyperparameters is done using grid search and numerical optimization. The results are evaluated according to the guidelines provided by the authors of the dataset, more specifically, root-mean-square error (RMSE) and a given evaluation function.An outcome of this work is the comparison between the methods considering both their evaluation score, and their training computational cost. The model is used to better predict the future condition of the engine and help assess the necessary actions that must be done for maintenance and repair in near and long-term future. This allows the optimization of the planification, which in its turn increases the availability. Importantly, it is possible to assess the condition of a component before the inspection, and thus focus on components that are predicted to be more degraded.&nbsp;

Data-Driven Remaining Useful Life Estimation for Predictive Maintenance of Gas Turbine Engines

No proceedings papers in 2021. Video only.

ASME 2021 Advanced Manufacturing & Repair for Gas Turbines (AMRGT2021)

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.