Data analysis techniques for the visualization and classification of historical vehicle engines' health status using data-driven solutions

In the field of cultural heritage, the use of non-destructive techniques to determine the state ofconservation of an artifact is of the utmost importance, to avoid damage to the object itself. In this paper, we present a data pipeline and several machine learning techniques for the visualization, analysis and characterization of engines in historical vehicles. The paper investigates the use ofvibro-acoustic signals acquired from the engines in different states of conservation and working conditions to train machine learning solutions. Data are classified according to their state of health and the presence of anomalies. The t-SNE algorithm is used for dimensionality reduction for data visualization. The machine learning algorithms tested showed encouraging performance in associating acoustic emission data with the engine signature, the type of anomaly and the working conditions. Nevertheless, a larger dataset would allow us to improve and strengthen the results.