Petronor – Promise, Improving Machine Reliability
Petronor
Sector: Industry
Business Case
Improvement of machine and process reliability through the application of models for defect diagnosis and prognosis.
Objectives
Increase device life and minimise unplanned downtime. Improve security.
Use case
Creation of diagnostic and prognostic algorithms. Design of an algorithm that measures asset health footprint and identifies asset defects and degradation.
Infrastructure
Hybrid On premise Cloud
Technology
Machine learning and deep learning
Data
Private dataset
Resources
The reliability, maintenance and system departments were involved. The deployed architecture occupies internal resources. The proof of concept lasted 2 years.
Difficulties and learning
The sensors used were not of the expected quality in terms of robustness, communication quality or accuracy. Sending the entire spectrum for further analysis and processing offers benefits in the quality of the algorithm, at the cost of increased reliability and bandwidth requirements in communications.
KPIs (business impact and metrics of the model)
The HW component is evaluated in terms of connection loss and packet shipment. The algorithms are qualified based on their ability to measure the quality of health footprint degradation and verify it in the in-plant context. Measuring assets’ RUL (remaining useful life).
Funding
Received fiscal aids for research projects in collaboration with agents of the BAC.
Collaborators, Partners
Tecnalia
