Welcome to my webpage. I specialize in applying data science and machine learning to heavy industries such as process, energy and chemicals where dealing with complex and messy sensor data is a challenge. In particular, I have a keen interest in topological and geometrical methods.
Here are some highlights of use cases for real-world challenges in heavy asset industries, focusing on enhancing efficiency, safety, and sustainability.

Selected Projects and Publications:

Topological Data Analysis for Condition Monitoring of Wind Turbines

Condition monitoring and predictive maintenance are crucial in data science as they help prevent equipment failures and reduce downtime, leading to significant cost savings and increased efficiency. In this project, TDA was applied alongside more traditional (Fourier) signal analysis, to condition-based monitoring (CBM) of wind turbines for energy generation. Data was acquired from a wind park in Norway using vibration sensors at different locations on the turbine’s gearbox, recording acceleration data and its frequency spectra at infrequent intervals. TDA was used to analyze the shape of the point cloud from vibration time series and generate topology-based key health indicators. These indicators were tested for CBM and fault detection, successfully identifying and classifying faults.
This work was presented at the 8th European Conference of the Prognostics and Health Management Society PHM-Europe 2024 and a full-text paper is available on the ArXiv or at the conference proceedings.

Hybrid ML Approaches to Virtual Flow Metering in Multi-phase Wells

Multiphase flow can be simulated and predicted in case sensor data are not available. While this could be done via either physical simulators or data-driven machine learning models, hybrid approaches provide a more versatile approach especially when coupled with cloud architectures and contextualised live data. This particular use case was dealing with wells at the end of their lifecycle when water cuts are high and accurate oil rates become economically more important.
This was published in Digital Chemical Engineering, Vol.9, page 100124 (2023)

Topological Data Analysis of Slug Flow in Offshore Wells

In this project, TDA (persistent homology) was used to perform signal analysis on offshore sensors data and condition monitoring of the multiphase flow. The undesired transition from regular to severe slugging flow is identified and classified with machine learning.
The article was published in Digital Chemical Engineering, Vol.4, page 100045 (2022) but there are a few typos in the formulae.
A corrected version of the article can be found at this link.

Gas-Surface Interaction in Materials: Hydrogen dynamics

In my academic days, I have worked at simulating the interaction between gas molecules and material surfaces. In particular, I worked on the dynamics of hydrogen on graphene and graphite surfaces.
Most of my work on this matter is summarized in a chapter part of this book.
Other studies were published in international scientific journals. You can find the whole list at my Google Scholar page.

Category Name 2

Page template forked from evanca