AI monitoring of wire drawing to reduce production downtime

The client faced high losses due to frequent breakdowns and uncontrolled deviations in the wire drawing process. The lack of timely anomaly detection led to frequent production outages, higher maintenance costs, and material waste. Traditional monitoring methods were unable to identify problems early or provide predictive insights, threatening production continuity and the ability to meet delivery deadlines. It was therefore crucial to deploy an effective system that could ensure early fault detection, minimize downtime, and increase overall production efficiency.

An AI system was developed to monitor the wire drawing process in real time. The solution uses machine learning to analyze sensor data and identify deviations from normal behavior. It automatically detects emerging anomalies, predicts possible failures, and alerts operators to risks before production breakdowns occur. The system is fully integrated into the production line, providing clear visualizations and maintenance recommendations. Thanks to a high degree of automation, it significantly reduces the need for manual inspection, supports predictive maintenance, and brings greater stability and efficiency to the entire production process.

production operators

6

Python, TensorFlow, PyTorch, scikit-learn, MQTT, OPC UA, SQL database, Docker, Grafana

Head of functional/operational unit

Full implementation

Joint management between client team and vendor.

Process engineers, production operators, IT/OT support

Industrial sensors and IoT gateways, server or cloud infrastructure for AI model execution, database storage for historical data, visualization dashboards, regular vendor support for model maintenance and retraining.

Quality improved significantly, scrap rates decreased, operator decision-making accelerated.

80% success rate in preventing product damage.

The client praised the customer-focused approach and emphasized that the solution prevents wire damage, reduces downtime, and saves material, energy, and labor.

Key success factors included close collaboration between client and vendor teams, high-quality and accessible production data, deployment of advanced AI models, and clearly defined business goals focused on minimizing downtime and material savings.

The biggest challenge was variability in production conditions and the need to secure sufficiently high-quality data for model training. This was overcome through intensive data collection, preprocessing, and gradual algorithm tuning in cooperation with process experts.

We recommend starting with a detailed analysis of production processes and data availability, involving process experts and end users from the start, and taking an iterative approach with rapid benefit validation. It is also crucial to prepare for change, ensure long-term AI model maintenance, and define goals and KPIs clearly at the beginning of the project to measure benefits and manage expectations effectively.