Cloud-based predictive maintenance and anomaly detection in textile production

Context

Textile machinery manufacturer, mechanical engineering. Complex mechanical systems in running production; machine data is meant to reveal failures early and at the same time support business reporting.

Problem

Emerging defects were detected late, causing downtime. Data collection was not designed for changing shift patterns, and daily production output could not be quantified cleanly.

What we built

A cloud-based predictive-maintenance solution: machine data from the factory is collected continuously in near real time and analysed for anomalies. Data collection was switched to the new shift structure (3 × 8 h → 2 × 12 h) for more accurate financial reporting, and daily yarn production is quantified — as the basis of a production-based pricing model.

Key engineering decision

Localise defects through deviations in performance behaviour — based on continuous collection rather than spot measurements. Analysing deviations from normal behaviour makes it possible to locate emerging defects quickly and precisely; the same data stream carries shift and production reporting.

Result

• +20 % operational efficiency
• Early detection of anomalies, reduced downtime
• More accurate financial reporting; quantification of yarn production as the basis for a production-based pricing model