Context and objectives
Frame and develop a data tool to improve a complex industrial process involving fermentation, in order to optimise both yield and output quality.
Outcome
- The product is in production and well adopted by process engineers.
- Average yield improvement so far: +1.2 percentage points.
- The product is now being rolled out to all factories and is becoming the reference, and its learning ability will further increase the uplift.
Our approach
Monitoring
- Real-time IoT data ingestion and batch monitoring
- Alerting
Model and interpretability
- A machine learning model with SHAP interpretability to understand multivariate correlations
Top and worst batches
- A rule induction engine that identifies the best and worst batches
Other case studies
All case studies- Supply chain modelling and optimisation Improving planning and logistics for agricultural commodity processing Read the case study →
- Augmented RAG for R&D Knowledge-graph-powered competitive intelligence on aircraft engines for a global aerospace leader Read the case study →
- Data innovation for quality improvement Valuing an in-house defect detection innovation for an advanced alloys manufacturer Read the case study →