Barilla uses machine learning for predictive maintenance and supply chain optimization

Overview

Barilla, the world's largest pasta producer, uses MLflow on the Databricks Data + AI Platform to operationalize machine learning models across its manufacturing operations. Predictive maintenance models forecast machine failures so maintenance can be scheduled proactively, reducing unplanned downtime, energy use, and emissions. Manufacturing cost deployment models help staff extract and allocate industrial costs and quantify plant losses. The company runs 90% of its business (finance, R&D, manufacturing, supply chain, marketing) on the platform, with 40+ data products in production and 1,000+ dashboards, saving millions of euros in production costs.

The challenge

Barilla's legacy on-premises data warehouse was ill-equipped to scale their operations effectively, creating silos that impacted their ability to navigate a growing list of external threats, and data analysts struggled to integrate and analyze the data needed, leading to delays and inefficiencies in delivering value to the business.

The solution

Barilla moved its data strategy to Azure Databricks, using Delta Lake to create a centralized global repository, Unity Catalog for data discovery and governance, Databricks Lakehouse for BI workloads, and MLflow to operationalize machine learning models; with manufacturing cost deployment models, staff extract and allocate industrial costs and quantify plant losses, and predictive maintenance models forecast when a machine is likely to fail so maintenance can be scheduled proactively.

Reported business value

By minimizing unplanned downtime, Barilla has been able to reduce operating costs with less energy and lower emissions associated with running equipment, and Databricks has helped Barilla solve business challenges across a range of departments, saving millions of euros in production costs.