Fedrigoni Industrial AI Challenge - Team 3

Project Overview

This project is part of the Fedrigoni Industrial AI Challenge where we focused on improving the slitting process in paper manufacturing at Fedrigoni’s Arco plant. By leveraging predictive AI models and data analytics, we aimed to predict daily slitting output and optimize production management to improve operational efficiency and customer satisfaction.

The goal of this project was to help Fedrigoni move from reactive to predictive production management, allowing them to anticipate low-output scenarios and make proactive adjustments to production workloads.

Objectives

• Analyze and identify key drivers of variability in the slitting process, such as machine performance, order profiles, and product mix.
• Develop predictive models to forecast daily slitting capacity based on various inputs.
• Improve operational efficiency by anticipating low-output scenarios and redistributing workloads to prevent delays and downtime.

Methodology

We followed the CRISP-DM methodology, which helped guide our data preparation, modeling, and evaluation processes:

1. Business Understanding: Defined the key operational problems in the slitting process.
2. Data Understanding: Collected and analyzed production data, including machine performance, order intake, and monthly production forecasts.
3. Data Preparation: Cleaned and integrated various data sources to make them suitable for modeling.
4. Modeling: Applied a range of ensemble forecasting models to predict daily slitting capacity.
5. Evaluation: Assessed the models using performance metrics to determine the most reliable solution.
6. Deployment: Discussed potential real-world deployment and integration within Fedrigoni’s production system.

Key Results

• Achieved a 95% accuracy rate in forecasting daily slitting output based on historical production data.
• Developed a predictive tool capable of simulating up to 3 days of production output, allowing the operations team to take proactive measures.
• Successfully maintained slitter utilization around 85%, improving production efficiency and lead times.
• Enabled the shift from reactive to proactive production management, helping to balance workloads and enhance customer satisfaction.

Technologies Used

• Programming Languages: Python
• Data Analysis: Pandas, NumPy
• Modeling & Forecasting: Ensemble models (e.g., Random Forest, XGBoost)
• Data Visualization: Matplotlib, Seaborn
• Machine Learning Libraries: scikit-learn, TensorFlow

Next Steps & Future Work

• Real-time Data Integration: The next step involves integrating real-time machine performance data to enhance prediction accuracy.
• Deployment: Future work will focus on deploying the model into a live production environment.
• Scalability: The model could be extended to other production lines within Fedrigoni, improving overall manufacturing operations.

Contributions

This project was completed by Team 3 for the Fedrigoni Industrial AI Challenge. Contributions included data analysis, model development, feature engineering, and evaluation.

Special thanks to Fedrigoni’s Arco plant for their collaboration and support in providing insights into the production environment.

Notes

• Confidentiality: Certain technical details, such as specific machine learning models, parameters, and proprietary data, have been omitted for confidentiality purposes.
• Data: All data used in this project is proprietary and provided by Fedrigoni for the purpose of this challenge.

Conclusion

This project demonstrates the potential of AI and data analytics to optimize industrial processes. By predicting slitting capacity and automating production management decisions, we can help companies like Fedrigoni increase operational efficiency and better serve their customers.