challenge
Our customer is a leading international healthcare group and develops, produces and sells pharmaceuticals worldwide. The challenge was that multivariate deviations from production parameters were only identified at the end of the production process in quality control. The result was unplanned production downtime and scrap costs due to poor quality. The aim was to implement predictive anomaly detection in real time using machine learning models and integrate them into the production monitoring system.
solution
- Preparation and preprocessing of sensor data for over 500 production process parameters
- Identification of significant parameters using statistical methods
- Machine learning model development for predictive real-time anomaly detection in Python
- Integration of the ML model into the production monitoring system
Business benefits
- Reduction of production downtimes
- Reduction of scrap costs due to poor production quality
- Real-time intervention in production control for quality optimization
