Trustworthy Predictive Maintenance (TPdM)

Purpose and goal

Predictive maintenance (PdM) has the highest potential to generate business value in the era of industrial digitalization. PdM solutions must be extended to provide interpretable results with increased accuracy in predictions through trustable decision support systems to achieve the vision of failure-free production. Therefore, the TPdM project aims to design human-centered decision support prototypes for PdM to achieve actionable decisions using advanced data science and scale up the innovative PdM applications in the Swedish manufacturing industry.

Expected effects and result

The expected results are identified models for roadmaps, models and methods for trustworthiness in PdM, designed and deployed software prototype for TPdM, and dissemination materials for spreading gained knowledge (e.g., lifelong learning). The impact of these results is efficient maintenance planning with reduced downtime, cost efficiency, increased OEE, productivity, robustness, resource efficiency, collaboration, and competence in smart maintenance as well as advanced data analysis for the competitiveness of the Swedish industry.

Planned approach and implementation

The planned approach for TPdM is linked to the project development structure through on-site/online software prototyping, with the iterative involvement of the decision makers using their experience in identifying machine wear patterns. This relies on interpretable data analysis based on multiple sources of information and modeling of uncertainties through machine learning, e.g. Bayesian statistical data analysis. These prototypes will enable the industry to make interpretable/trustable decisions when implementing PdM solutions.