AI4Production – Real-Time Anomaly Detection for Electronics Assembly

Use Case:

Automated analysis of sensor data during electronics production to detect anomalies and reduce downtime and faulty assemblies with integration into Azure cloud and production line software.

Outcome:

• Accuracy of 98.87% in detecting electronics with faulty wires
• Full automation of weekly and monthly anomaly detection tasks across 14 monitored parameters
• 0 false negatives in detecting faulty wires—ensuring critical defects are always caught
• 40% reduction in time and cost for production engineers’ monitoring activities
• Email notification system with visual insights and alerts on trend deviations and defects

Ecosystem Support:

Supported by StairwAI expert matchmaking and advisory.

AI Relevance:

1. Demonstrates how domain SMEs can leverage AI for real-time operational intelligence.
2. Combines statistical and ML models for interpretable and actionable monitoring
3. Edge-to-cloud deployability using Docker and REST APIs with Heroku & Azure integration.
4. Enables full traceability and early defect detection for electronic product lines

Summary:

Senso4s, a Slovenian SME specialising in smart sensors, tackled a high-cost issue in electronics assembly: undetected faulty wires in final products. With StairwAI’s support, they developed AI4Production, a comprehensive solution combining cloud telemetry, statistical modeling, and anomaly detection. The system continuously monitors 14 electrical parameters collected from each product under test (e.g., battery voltage, BLE signal strength, supply currents), and uses a combination of Isolation Forest, Matrix Profile, and Prophet to flag outliers and trend changes. Statistical tests like the Augmented Dickey–Fuller (ADF) and Mann-Kendall Trend (MKT) ensure anomalies are statistically robust before triggering alerts.

To address imbalanced datasets (e.g., very few defective samples), the team used synthetic data generation and oversampling techniques alongside Gradient Boosted Trees, reaching 100% recall on defective wires with no false negatives. An automated email system sends engineers clear, visual alerts whenever anomalies are detected, complete with trend plots and defect details. The system is fully integrated into Senso4s’ Azure-based infrastructure via Dockerized containers. One container monitors incoming data and schedules detection; another runs the computations and alerts, enabling a hands-free quality control loop. Weekly and monthly reports allow engineers to act on detected process shifts—cutting waste, boosting uptime, and improving production insights. By integrating AI into their quality pipeline, Senso4s has not only reduced internal effort but also laid the foundation to commercialize AI4Production for external electronics manufacturers, particularly in the growing global consumer electronics market. This project is a compelling example of how SMEs can deploy intelligent analytics with minimal infrastructure change to unlock immediate ROI and long-term process visibility.