About
Dr
Vivek Mahato
Postdoctoral Researcher
Dr. Vivek Mahato is a Postdoctoral Researcher in I-Form, working in the area of Machine Learning. He obtained his undergraduate degree in Computer Applications from WBUT, India. Furthermore, his interest in machine learning and research drove his ambition to attain a Master’s and then a PhD in Data Science at UCD. His main research interests are data analytics, machine learning, time series analysis, recommender systems, and their practical solutions for the manufacturing sector.
Research Interests (Lay Summary)
In the age of Industry 4.0 (the fourth industrial revolution, with a growing trend of utilising cyber-physical systems in manufacturing), industries are capitalising on the opportunity to use advanced machine learning (ML). Using ML algorithms, the idea is to make manufacturing processes smarter and maximise environmental sustainability, health and safety, and economic competitiveness. Furthermore, the manufacturing industry is experiencing an unfathomable rise of accessible data, and we are witnessing a paradigm shift from basic statistical models to pattern recognition. Dr. Mahato's research interest lies in having real-time predictive control over the Additive Manufacturing process (an urgent requirement). A real-time system flags an anomalous event happening in production, alerts the operator or engineer, and recommends a set of build parameters to tune. Having such a system thus significantly reduces overhead costs, material wastage, and energy consumption, while also achieving better quality products.
Technical Summary
Dr. Mahato’s research focuses on how to bring Machine Learning techniques into the manufacturing domain. He considers a particular category of time-series analysis, which is typical in a production environment. However, the current literature presents a limitation of readily available tools to address this problem. Therefore, he intends to utilise data generated during the process of manufacturing to extract possible patterns and anomalies for exploratory analysis. The current state-of-the-art is using agglomerative statistical metrics to summarise the data for the study, but he is motivated to use the raw time-series data directly, for he believes that such summaries discard valuable insights from the data that is required for a cogent analysis or monitoring of the process.