PhD Position in Trustworthy Artificial Intelligence for Predictive Maintenance of Industrial Equipment at Politecnico di Milano
We are pleased to inform you that POLITECNICO DI MILANO (Laboratory of Analysis of Systems for the Assessment of Reliability, Risk and Resilience (LASAR3) www.lasar.polimi.it) are accepting candidatures for a PhD position on
TRUSTWORTHY ARTIFICIAL INTELLIGENCE FOR PREDICTIVE MAINTENANCE OF INDUSTRIAL EQUIPMENT
Location: Laboratory of Analysis of Systems for the Assessment of Reliability, Risk and Resilience (LASAR3), Politecnico di Milano
Website: www.lasar.polimi.it
Application Deadline: December 10, 2024
Grant Starting Date: December 13, 2024
Application Link: Online Application
Call Description: Detailed Call Information
(After registering, you can use your Person Code and password to access the PhD application: online services - Application forms - Application for "bandi aggiuntivi" PhD programmes)
Motivation and objectives of the research in this field:
The increased availability of data from industrial equipment and the grown ability to treat these data by artificial intelligence (AI) methods have opened the doors for the development and application of predictive maintenance in several industrial sectors, like nuclear, oil and gas, energy, electronics and transportation. Predictive maintenance is a great opportunity for reducing unforeseen failures, and operation and maintenance costs, and for increasing components and machines usage. Practical implementation of predictive maintenance entails trustworthiness of the AI model outcomes. This brings some key challenges of AI: i) data collected from degraded and failed equipment are scarce; ii) AI models must capture the inherent uncertainty of the degradation and failure processes; and iii) the AI models are of black-box nature and their outcomes are difficult to explain physically. The research of the proposed PhD thesis aims at developing new AI analytics for predictive maintenance based on the powerful emerging techniques of Deep and Transfer Learning, Physics-Informed Neural Networks and Generative Adversarial Networks for improving prediction accuracy, especially in case of imbalanced data, i.e. dew degradation and failure data (challenge i); Deep Ensembles and Bayesian Neural Networks for treating the uncertainty in AI models (challenge ii) and eXplainable Artificial Intelligence algorithms for enhancing model transparency and explaining AI model outcomes (challenge iii).
Methods and techniques that will be developed and used to carry out the research:
Artificial Neural Networks, Bayesian Neural Networks, Physics-Informed Neural Networks, Deep Ensembles, Autoencoders, Convolutional Neural Networks, Generative Adversarial Networks, Transfer Learning, eXplainable Artificial Intelligence.
We invite all interested candidates to apply and share this opportunity widely. For further details or queries, please feel free to reach out to:
Ibrahim Ahmed, Piero Baraldi, Francesco Di Maio, Enrico Zio