Anomaly detection in industrial processes

Purpose and goal:

The scope of this project was to identify novel areas where the expertise of MIUN and Ontario Tech University can be applied to enhance industrial processes for the partner companies, SCA and Pandrol. From a combination of meetings and workshops, we identified relevant industrial challenges that could be elaborated into novel methodologies that improve the industrial operations, but also have scientific relevance.

Both industrial challenges provided by the companies fitted well under the anomaly detection domain, focusing on two perspectives, time series and image-based. Expected results and effects:

From the scientific perspective, the main outcomes of this project are the three scientific publications in industrial anomaly detection, where two are in collaboration with SCA and one in collaboration with Pandrol. Aside from them, important outcomes are the future projects that were identified with both SCA and Pandrol. Due to change of ownership at Pandrol, we will not pursue one of the projects.

The project we aim to pursue is a collaboration between SCA, MIUN and Ontario Tech University for the prediction of lifetime in bearings. Approach and implementation:

Lifetime prediction in bearings can be crucial in industrial processes because it enables planning of preventive maintenance, optimises costs, and ensures reliability and durability. From the research perspective, predicting lifetime from factory metadata in a noisy factory environment raises important challenges. For this reason, we aim to develop a project new project involving SCA, MIUN and Ontario Tech University. The scope will be to develop methods that facilitate accurate lifetime prediction.