Future AI-based maintenance

Purpose and goal

The goal of this project is to develop a general framework to facilitate the use of advanced predictive maintenance and planning under uncertainty in industry. The project aims also to reduce business risks by switching to more efficient and AI-based maintenance strategies and thereby increasing resource efficiency and competitiveness in Swedish industry.

 The project develops the following:

1. A modular platform for prediction models - Where existing and novel prediction models for component wear can be combined, and incrementally refined over time.
2. New and improved component lifetime prediction methods - Based on machine learning and causal inference, taking uncertainty into account.
3. New methods for dynamic and robust maintenance planning under uncertainty.
4. A simulation and evaluation tool - That can be used for decision support to evaluate and compare advanced maintenance strategies regarding risk and cost. 

The tool for comparing and evaluating the potential of different maintenance strategies will help to significantly decrease the business risk involved in changing from an old but historically proven maintenance strategy into a more efficient one. Therefore, it will enable a transition to more efficient and AI-based maintenance strategies in industry, which will have a huge impact on the resource efficiency and competitiveness of industry at large.

Implementation

The project is implemented through five work packages (WP), described here below:

• WP1 - Project management and knowledge transfer
  • Lead: RISE. Aim: Effective project organization.
• WP2 - Probabilistic modular platform for model integration  
  • Lead: RISE. Aim: Develop a modular platform based on a common probabilistic representation of lifetime predictions, costs, and risks, that will be used as a glue between both data driven and simulator-based models for different components, and the optimization algorithms
• WP3 - Data-driven causal estimation of remaining useful life with uncertainty measure
  • Lead: Scania CV. Aim: Build data-driven wear models, for prediction of remaining useful life, including uncertainty measures and causal inference.
• WP4 - Dynamic robust maintenance planning under uncertainty 
  • Lead: Siemens Energy. Aim: Develop an optimization model for maintenance planning that considers the predicted remaining life of components as well as business requirements
• WP5 - Maintenance strategy evaluator
  • Lead: RISE. Aim: To develop a tool for simulating maintenance strategies, with the purpose of comparing and assessing the gains of different maintenance strategies.

Budget and participants

The project budget accounts to 11.18 million SEK, of which 5.59 million SEK is financed by Vinnova, as part of the call Advanced and innovative digitalization 2023 - call two. The main coordinator is RISE – Research Institutes of Sweden, and project partners are: Scania CV and Siemens Energy. The consortium partners have experience in both applied and basic research on predictive maintenance as well as ML and AI for over two decades.