ReMAIntAIn

Full Title: AI-Based Repair and Reuse in Resource Intensive Machinery

Duration: 01/04/2025 - 31/03/2028

Assignment to our (IMW) research priorities:

  • Prescriptive Maintenance
  • Circular Economy
  • Production Optimization
  • AI for Industry
  • Natural Language Processing

Abstract:

ReMAIntAIn proposes a recommendation-based maintenance system for aluminum production environments. It integrates time series sensor data, textual maintenance records, and physical domain knowledge to enable adaptive failure prediction and support data-driven maintenance decisions. By leveraging non-stationary Bayesian networks and large language models, ReMAIntAIn estimates the Remaining Useful Life (RUL) of components under dynamic conditions. This facilitates optimized scheduling of maintenance actions and enhances component repairability and reusability.

Results:

  • A non-stationary dynamic Bayesian network will be developed to model the evolving temporal dependencies in a dynamic industrial environment. This model will capture changes in system behavior over time, improving maintenance prediction, supporting Remaining Useful Life (RUL) estimation, and optimizing maintenance decision-making to enhance repairability and reusability.
  • An LLM-based solution is proposed to analyze and extract valuable insights from object-oriented and order-oriented historical data. This will enhance the understanding of maintenance patterns, enabling better RUL predictions, and improving repairability throughout the maintenance process. 

Partners:

Contact Details:

Univ.-Prof. Dr.-Ing. habil. Fazel Ansari

Email: fazel.ansari@tuwien.ac.at