Overview
The maintenance of industrial equipment is typically preventive, scheduled based on calendar time or asset runtime, or reactive when errors occur. Today’s availability of huge amounts of sensor data permits predictive maintenance, which is scheduled as-needed based on real-time conditions of industrial assets. Predictive maintenance tracks the performance of equipment during normal operation and detects possible defects before a failure occurs. This maintenance strategy has two major advantages: cost savings due to a reduction of downtime since maintenance steps can be better planned, and a reduction of resources since parts are only changed if their performance degrades.
The PREMISE project aims at developing a framework for predictive maintenance for industrial assets and is composed of three main components: data preprocessing, data analysis, and prediction. Since raw sensor data is typically noisy and incomplete, a preprocessing phase is needed to clean and prepare the data, before more advanced analysis steps can be applied. The analysis phase aims at extracting correlations between different signals, performing a root-cause analysis, and identifying signals and patterns with high prognostic value. The extracted knowledge serves in the prediction phase as a basis to forecast abnormal behavior and failures in a reliable manner.
The project is a collaboration between unibz and the two industrial partners Durst and TechnoAlpin, which produce digital printing systems and snow making systems, respectively. The developed prediction framework will be applied and evaluated using different use cases provided by the two companies.
Partners
