Work

Challenging the Dynamics of Time: Generate and Evaluate Real-World Time Series to estimate NOx Emissions in a Turbo Machine

Year: 2023
Type: dissertation
Abstract: Estimating accurate NOx emissions is essential to monitor pollution and health condition \nof a turbo machine. We use a virtual sensor to correctly estimate the particle pollution \nvalue in real-time, leveraging modern machine learning approaches. It is well known \nthat machine learning heavily relies on data, but real-world applications encounter datarelated \nissues. In this work, we address the cited business use case scenario where limited \ndata hamper an optimal regression quality. We investigate the application of time series \ngenerative models, with a particular emphasis on evaluating their performance using \na comprehensive set of quantitative and qualitative metrics. We also conduct a critical \nanalysis of the evaluation metrics commonly employed in the literature for validating and \nassessing the effectiveness of generative models. The analysis highlights the limitations \nof these metrics, as they do not take into account the temporal dependencies present in \ntime series data and rely heavily on the specific implementation of the evaluation model. \nFinally, task-specific metrics are proposed to assess the effectiveness of generated data \nin supporting an industrial application. By delving into those pillars, this work aims \nto contribute to the advancement of knowledge on temporal synthetic data generation, \nshowing how it can impact environmental care.
Language: English
Open Access status: green
OpenAlex ID: W7008520806
Cites: 29
Cited by:
Related to: 10