Ph.D. candidate in Ingegneria Meccanica , 40th cycle (2024-2027)
Department of Mechanical and Aerospace Engineering (DIMEAS)
Profile
PhD
Research topic
AI and Digital Twin Implementation in Steel Ball Manufacturing
Tutors
- Raffaella Sesana
- Irene Pessolano Filos
Keywords
Biography
The research will follow an iterative and data-centered methodology. Starting by establishing a trustworthy dataset by improving data acquisition and organization. Once the data is sufficiently structured, they will be analyzed using exploratory statistics and feature engineering to identify the most relevant input-output relationships. Then developing and comparing different AI models. Supervised machine learning methods will be tested first, followed by more advanced approaches such as artificial neural networks, depending on the characteristics of the data and process behavior. The final goal will be to identify a model capable of predicting or recommending lapping conditions that will satisfy quality constraints while minimizing processing time.
This phased approach will ensure that the research remains realistic, technically sound, and aligned with the needs of the industrial partner.
The expected contribution of the project will be both scientific and industrial. It will provide a case study on how AI methods can be applied to a complex precision manufacturing process where data quality will be uneven and process behavior will not be immediately easy to model. It may also contribute to the understanding of how data-driven methods can be combined in industrial optimization problems.
From an industrial perspective, the project will be expected to deliver a practical decision-support tool for the lapping process, with the potential to reduce cycle time, improve quality consistency, and support more efficient parameter selection. The outcome should also improve the factory's data collection practices by replacing or complementing paper-based recording with a more reliable digital workflow. The publication plan will be to produce two to three journal papers, preferably targeting Q1 and Q2 journals.
The training path will be designed to support the interdisciplinary nature of the project. It will include courses in scientific writing, Python programming, machine learning, data mining, optimization, digital twin modeling, and structural-integrity and AI-driven industrial decision-making topics. In addition, continuous improvement of Python, AI, and ML skills will be pursued through online courses. This training will be directly relevant to the project because it will strengthen both the technical and communication competences required for advanced doctoral research.
The first phase of the PhD will therefore focus on problem definition, literature review, industrial familiarization, and preliminary data collection. The next phase will focus on cleaning and structuring the lapping dataset, improving digital recording, and identifying the most promising AI approach. By the end of the current research stage, it will be aimed to have a successful model for lapping that can be validated in the industrial context and presented in a first journal paper. The final stage of the PhD will focus on consolidating the results, preparing the remaining publications, and integrating the findings into the doctoral dissertation.